This document is intended to be a scaffold for community discourse and creation of knowledge. The version hosted on the PubPub platform (find it here, if you aren't already viewing it) allows inline commenting. We invite ideas, additional references, and constructive criticisms via inline comments. We request that citations be included to support claims made in comments when possible. We intend to publish updates to this roadmap over time, which is possible on the PubPub platform. Well-referenced comments will be incorporated into future versions of the roadmap.
List of abbreviations
CA carbonic anhydrase
CapEx capital expenditure
CDR carbon dioxide removal
BECCS bio-energy carbon capture and storage
DAC direct air capture
DOC direct ocean capture
DIC dissolved inorganic carbon
MEA monoethanolamine
OpEx operating expenditure
TEA techno-economic analysis
Summary
This work focuses on methods that leverage the enzyme carbonic anhydrase (CA) or its mimics to reduce the energy requirements and cost for removing CO2 from the atmosphere using aqueous solutions. We consider methods for direct air capture (DAC), direct ocean capture (DOC), and point-source capture of biogenic CO2 (such as bio-energy carbon capture and storage (BECCS)) using aqueous solvents.
Methods for DAC and point-source capture using aqueous solvents are limited by a trade-off between the energetics of solvent regeneration and the kinetics of absorption of CO2, which translates to a trade-off between energy usage (a component of OpEx) and CapEx of the air contactor. CO2 hydration is the rate-limiting step determining CO2 absorption kinetics.
The trade-off may be addressable using CA, which catalyzes CO2 (de)hydration. Efficient application of CA during CO2 absorption could enable fast absorption in solvents with low regeneration energies, resulting in substantial cost and energy savings.
Current DOC methods use seawater acidification to increase the rate of CO2 outgassing. This approach requires using an electrolyzer for acid generation, which is a leading cost factor.
The rate of CO2 outgassing may be enhanced by using CA or its mimics to catalyze bicarbonate dehydration with less or zero acidification, reducing the energy demands of electrolysis or obviating the electrolyzer.
In both DAC and point-source capture, challenges to beneficial use of CA or its mimics involve developing CA proteins and CA-bearing composite materials that are active near the air-solvent interface, active under process conditions, and economical; and avoidance of biofouling.
In DOC, challenges to beneficial use of CA or its mimics involve developing hollow fiber membranes containing the catalysts that are long-term stable, have low membrane resistance to mass transport of CO2, are low cost, and avoid biofouling.
For DAC, DOC, and point-source capture, process modeling and techno-economic analysis (TEA) should be used to guide process designs and provide performance targets for enzymes and enzyme-laden composite materials. In modeling, the properties of CA and its immobilization format should be considered mutable (i.e., engineerable).
A wide variety of naturally-occurring forms of CA exist in nature, with considerably different stability and activity with respect to temperature, pH, alkalinity, and salinity. Additional forms have been artificially engineered toward applications in point-source capture, with early modeling indicating cost-competitive performance. Ongoing, rapid progress in protein engineering, particularly in emerging methods leveraging machine learning, as well as "bio-mining" from metagenomic sequences, suggests that producing highly process-optimized CA is increasingly feasible.
Prize-based competitions could be used to direct protein engineering efforts toward clearly-defined performance goals for CA.
All solutions considered in face the challenge of establishing and supporting interdisciplinary teams that are optimally suited to the engineering challenges.
We recommend actions to advance the above-mentioned methods. Most recommendations relate to all of the methods simultaneously.
Roadmap structure
Key concepts: We first introduce key CDR technologies and underlying concepts (Section 1)
Constraints: We describe constraints on CDR technologies and divide them into categories of cost, scaling, and environmental. We focus on constraints that may be addressable by biotechnology, rather than giving a complete list of constraints on each CDR technology (Section 2)
Possible solutions: We identify possible solutions through which biotechnology could alleviate the constraints (Section 3, with specific solutions in Sections 3.1, 3.2, and 3.3).
Challenges: We identify challenges in implementing the solutions (Sections 3.1.3, 3.2.3, and 3.3.3) and describe their potential benefits in terms of the constraints.
Recommendations: At the end of the roadmap, we provide a list of recommendations for research directions that collectively enable possible biotechnology solutions to constraints on CDR technologies (Section 4). Table 10 summarizes the relationship between recommendations, possible solutions, constraints, and CDR technologies.
1. Introduction to industrial capture of CO2 using aqueous solutions
This work focuses on methods that leverage enzymes to reduce the energy requirements and cost for removing CO2 from the atmosphere using aqueous solutions1. Specifically, we explore using the enzyme carbonic anhydrase (CA) to catalyze CO2 hydration and dehydration in direct air capture (DAC), point-source capture for bioenergy carbon capture and storage (BECCS), and electrochemical direct ocean capture (DOC). As discussed below, the DAC and point-source BECCS capture methods we consider use circular cycling of an aqueous solvent with absorption and degassing of CO2 within an industrial facility. In contrast, electrochemical DOC degasses CO2 from a continuous input stream of ocean water within an industrial facility, while absorption of CO2 from air occurs in the open ocean.
In general, the advantage of using CA is a possible reduction in energy requirements by catalyzing hydration and dehydration of CO2 in aqueous solution, which are the rate-limiting kinetic steps in CO2 absorption and outgassing, respectively. As discussed below, catalyzing CO2 hydration and dehydration in DAC and point-source capture could alleviate the tradeoff between absorption kinetics and regeneration energy that is typical of aqueous solvents by enabling the use of solvents that have relatively low regeneration energies but have unacceptably slow absorption kinetics in the absence of the catalyst. Similarly, catalyzing CO2 dehydration in DOC can reduce cost by increasing the rate of CO2 outgassing from a given quantity of degassing equipment.
1.1 Brief overview of direct air capture and point-source capture of CO2
Direct air capture (DAC) is a leading form of CDR that involves processing ambient air in industrial facilities to remove CO2(Erans et al., 2022), typically producing high-purity CO2 (e.g., > 95%) that is ready for storage. In many ways, DAC is an ideal form of CDR. Relative to other methods, DAC is easily measurable, has flexible siting, and can connect with durable storage methods (notably geologic storage). Additionally, DAC is insensitive to future climate change-induced changes such as increased wildfire risk, increased mean temperature, and extreme weather, which threaten methods based on biomass cultivation, forests, and soils. Finally, scientific uncertainties and risks around the impacts of carbon storage in direct contact with ecosystems, such as soils and seawater, are less relevant for DAC, whose principal technical challenges and scientific unknowns are all engineering-related. However, DAC has the key disadvantages of high cost (estimates range from $100 to $1000/tCO2(Erans et al., 2022) and high energy consumption relative to other CDR methods. As such, improvements in the efficiency of DAC that both decrease energetic requirements and bring it within a competitive cost range could have profound implications for the feasibility of durable carbon removal with minimal environmental risk.
“Point-source capture”, or sometimes simply “carbon capture,” refers to a set of technologies used to separate CO2 from industrial gas streams before they are vented to the atmosphere. These gas streams, such as flue gas from fossil fuel combustion or concrete manufacture, typically contain a mixture of gasses (commonly N2 and H2O) with CO2 present at ~5-30% by volume, much higher than atmospheric concentration (currently ~420 ppm, or ~0.042% by volume). CO2 captured from point sources can also be stored geologically.
Point-source capture can amount to atmospheric CDR in the case of BECCS, wherein biomass fixed from atmospheric CO2 (e.g., plant biomass) is combusted, pyrolyzed, or gasified to produce energy or fuel and a waste gas stream containing CO2 that can be separated using point-source capture technology and then sequestered (Fig. 1).
Figure 1. Categorization of methods for separating CO2 from mixed gas. Atmospheric carbon dioxide removal is achieved when CO2 is removed directly from the atmosphere, or from an industrial gas stream derived from biomass.
In both DAC and point-source capture, CO2 is separated from a gas mixture by repeatedly cycling a process of 1) exposing the gas mixture to a capture material with selective affinity for CO2 until it is loaded with CO2[2], (Bui et al., 2018); and 2) generating a “parameter swing” (e.g., a change in temperature, pressure, or pH) that causes the CO2 to be released at high purity for collection and regenerates the capture material for reuse (Fig. 2). A variety of capture materials are in use or under development for both DAC and point-source capture (Sanz-Pérez et al. 2016; Bui et al., 2018; Küng et al., 2023).
We will focus on aqueous liquid solvents because their limitations are, by our assessment, addressable by biotechnology. CO2 hydration in the aqueous solvent is the rate-limiting step of CO2 absorption. In turn, CO2 absorption rate strongly influences CapEx and energy usage. As a catalyst of CO2 hydration, CA therefore can offer substantial benefit, as discussed in detail below. The use of aqueous solvents in DAC and point-source capture is also already relatively technologically advanced , and aqueous solvents have the advantage of being cheap, long-lived, and non-toxic relative to other capture materials.
Figure 2. High-level schematic of separation of CO2 from mixed gas using aqueous solvent. Air or flue gas is exposed to a solvent in a contactor, after which CO2-depleted gas is discharged. The CO2-loaded solvent then flows through a regeneration process, where CO2 and lean solvent are regenerated. Various absorption and regeneration processes are used for different implementations of DAC and point-source capture. In DOC, the ocean surface is effectively the contactor, and ocean water that is equilibrated with atmospheric CO2 is effectively the loaded solvent.
Fig. 3 shows a breakdown of the processes involved in solvent-based capture. To absorb CO2 into an aqueous solvent, air or flue gas is moved over a high-surface-area packing material over which the solvent flows, similar to a cooling tower. Fundamentally, these solvents leverage an alkaline solution to absorb CO2 in the form of (bi)carbonates. For solvent regeneration in DAC, the most prominent aqueous solvent regeneration methods are the calcium looping system of the company Carbon Engineering (Keith et al., 2018) and electrochemically-driven pH swings (Sharifian et al., 2021), used, for example, by the company Mission Zero. Alkalinity swing has also been proposed for regeneration, although it is not implemented commercially (Rinberg et al., 2021). For regeneration in point-source capture, thermal swing is most commonly used commercially (Bui et al., 2018).
Figure 3. Breakdown of the process parameters contributing to CapEx and OpEx in solvent-based capture of CO2 (reproduced from [9] with reformatting, with permission from Springer Science+Business Media, LLC).
The three primary performance characteristics influencing the choice and engineering of the capture material are the kinetics of loading CO2 in/onto the material, the energy required for regeneration, and the loading capacity. Faster kinetics increases the sorption rate of CO2, which allows for a smaller contactor tower because the contactor size is inversely related to kinetics of CO2 capture for a given CO2 capture rate (Fig. 4). This is significant because the air contactor is a large fraction of the system CapEx. Kinetics also influences the inventory (total amount) of solvent that must be present in the system to enable sufficient dwell time (this is also influenced by solvent loading capacity, see below). The energy for regeneration is the main determinant of the total energy required to capture a target quantity of CO2. Energy is the primary component of OpEx. As noted, loading capacity influences the total quantity of capture material needed and/or the cycling speed required to capture a target quantity of CO2, which contributes to cost through materials usage or cycling (e.g., pumping of an aqueous solvent). Therefore, the search for a cheap, durable, fast-absorbing capture material with high loading capacity and low regeneration energy is a significant research focus. Modeling can be used to identify the optimal sorbent properties for a given process design, informing the choice and engineering of sorbents (Young et al., 2023).
For aqueous sorbents, the key challenge is overcoming a tradeoff between kinetics of absorption and regeneration energy, as discussed below.
Figure 4. The size of the contactor unit for DAC and point-source capture scales as the inverse of absorption kinetics, for a given CO2 capture rate. The plot shows contactor size and absorption kinetics, normalized to an arbitrary set of values for size and kinetics for a given overall CO2 capture rate.
1.2 Brief overview of direct ocean capture of CO2
DOC achieves CDR indirectly by using an industrial facility to remove carbonates from ocean water in the form of CO2 (de Lannoy et al., 2018; Digdaya et al., 2020).2 The captured CO2 is sequestered using methods similar to DAC and point-source capture, while the CO2-depleted ocean water, which has an elevated pH, is released back into the ocean. Equilibration of air with the resulting dissolved inorganic carbon (DIC)-depleted ocean water removes CO2 from the atmosphere (Fig. 5A). In a sense, DOC can be conceptualized as a DAC process in which ocean water is the aqueous solvent and the ocean surface is the air contactor (Fig. 2). Because DOC leaves air capture to the ocean itself, the technical challenges of DOC only involve removing CO2 from solution.
From a technical standpoint, the process of removing carbonate ions from ocean water is similar to the CO2 regeneration step in DAC and point-source capture using aqueous solvents: all three processes remove dissolved carbonate ions from an aqueous solution to produce a concentrated stream of CO2. However, the CO2 loading capacity of ocean water is ~500-fold smaller than that of the aqueous solvents used in DAC and point-source capture.3 That means DOC has a lower rate of CO2 outgassing and thus requires a large surface area for water degassing and acidification to maximize outgassing rates, which in practice means a high-CapEx apparatus.4 The kinetics of removing CO2 from ocean water are therefore a crucial factor in DOC: they control the outgassing rate and thus the CapEx. Leading proposed methods for DOC operate using a combined pressure swing and electrochemically-driven pH swing (de Lannoy et al., 2018; Digdaya et al., 2020; Kim et al., 2023), as discussed further below.
Figure 5.A) The DOC facility extracts CO2 from seawater using a combined pH and pressure swing and returns DIC-depleted seawater to the ocean. CDR is achieved as atmospheric CO2 enters the ocean to equilibrate with the DIC-depleted water, and the removed CO2 is transported away for storage. B) Within the DOC facility, seawater enters a bipolar membrane electrodialyzer and is acidified (pH swing), after which it is passed through hollow fiber membranes under vacuum pressure for CO2 degassing (vacuum swing). The DIC-depleted seawater, still acidified, flows back into the electrodialyzer to be basified. It exits the facility with a slightly higher pH than it entered, due to removal of CO2.
1.3 Absorption and regeneration of CO2 in aqueous solutions
To approach CDR using aqueous solutions, it is important to understand how CO2 moves from air or flue gas into water and the resulting carbonate chemistry. During absorption of CO2, gaseous CO2 (CO2(g)) dissolves into water to become an aqueous dissolved gas (CO2(aq)), at a rate proportional to the difference in the partial pressures of CO2 in the gas and the solution relative to their equilibrium partial pressures. CO2(aq) then becomes hydrated and speciated into carbonic acid and (bi)carbonates (Fig. 6A). The reverse processes occur during outgassing. The dominant parameter influencing the distribution of these forms of inorganic carbon is pH, with CO2 dissolving into solution at high pH and outgassing at low pH (Fig. 6B).
Figure 6. A) Chemical equations of CO2 dissolving and speciating, including (de)hydration kinetics (Sharifian et al., 2021). B) Bjerrum plot describing the fraction of DIC accounted for by each species as a function of pH. For interpretation, note that the total DIC increases with pH, meaning that, for example, there is more bicarbonate dissolved in solution at its pka2 than at its pka1, even though bicarbonate accounts for ~50% of DIC in both cases. (reproduced from (Zosel et al., 2011) with reformatting and annotation with pKa values, with permission from IOP Publishing, Ltd.)
As discussed above, the three key features of an aqueous solvent for DAC or point-source capture are the rate at which it absorbs CO2, its regeneration energy, and its capacity to hold CO2. For DOC, the capture solvent is ocean water, which cannot be modulated.
The capacity for an aqueous solvent to absorb CO2 depends on its alkalinity, i.e., its ability to balance the acidity of CO2 by accepting protons. Thus, strongly alkaline solvents have high CO2 capacity and are frequently used as capture materials. Options for strongly alkaline solvents include high concentrations of either strong or weak bases. Strong and weak bases can be equally alkaline,5 but they have opposing advantages in terms of absorption rate and regeneration energy (Table 1).
Weak base
Strong base
Example solution
1M K2CO3, pH 9
1M NaOH
pH
9
14
alkalinity
1 eq/L
1 eq/L
CO2 regeneration energy
low*
high*
normalized hydration rate**
10-5
1
Table 1. Comparison of properties of a weak base and strong base of equal concentration. Note that alkalinity is equal, meaning the capacity to absorb CO2 is equal. *For a sense of regeneration energy, Keith et al. 2018 (Keith, et al. 2018) states that regeneration by calcining, which requires less energy than regeneration from 1M NaOH, required 4.05 GJ/tCO2 (178.3 kJ/mol), while (Reardon, 2015) found that regeneration from a salt solution (possible K2CO3) required 2.11 GJ/tCO2. **Absolute hydration rate depends on the concentration of CO2, so we have given the normalized rate relative to 1M NaOH.
The CO2 absorption rate depends on the kinetics of the equations in Fig. 6A. All of the rate constants are fast (practically instantaneous) except for the (de)hydration rate constants, k2, k-2, k3 and k-3, which are orders of magnitude slower. Hydration kinetics therefore determine the absorption rate, and as a result they are a central focus of this work as mentioned above. Above pH ~8.7, the concentration of OH- is high enough that the rate of Eqn 3 dominates over Eqn 2 (Fig. 6A). Solvents with high OH- concentration (strong bases with high pH) are thus most favorable for absorption kinetics, while weak bases, which have lower pH, are less favorable.
Regeneration energy—the energy required to remove CO2 from the solvent—depends on the enthalpy of absorption of CO2 in the solvent. Because CO2 is acidic, it has a higher heat of absorption in strong bases than in weak bases. Strongly alkaline solvents composed of weak bases with moderate pH (i.e., high concentration of weak base but lower OH- concentration) are therefore most favorable for regeneration energy.
The competing benefits of strong and weak bases as capture solvents result in a tradeoff between absorption kinetics and regeneration energy in DAC and point-source capture. Resolving the tension between absorption kinetics and regeneration energy in aqueous solvents is therefore a substantial technology-development focus.
1.4 Aqueous solvents in DAC and point source capture
In practice, contactor size dominates overall costs, making fast absorption kinetics highly desirable, even at the cost of higher regeneration energy, because fast kinetics allow more CO2 absorption per unit area. Strong bases with high absorption kinetics and high regeneration energies, such as potassium hydroxide (KOH), are therefore the most prominent aqueous solvents used in DAC.
Calcium looping and pH swing are currently the predominant aqueous solvent-based DAC methods. For example, Carbon Engineering’s calcium looping method uses a KOH solvent of pH ~14, and then precipitates the resulting dissolved carbonates into solid calcium carbonates. The solid calcium carbonates are subsequently heated to 900 ˚C to release pure CO2(Keith, et al. 2018). pH swing methods typically use a swing from pH ~2.5 to pH 13 (Eisaman et al., 2011; Sabatino et al., 2020). While pH swing methods are diverse (Sharifian et al.,2021), a conceptually simple example leverages bipolar membrane electrodialysis to dissociate water into acid and base streams. The base stream is used as a CO2 capture solvent and the resulting (bi)carbonates are electrochemically transported across an anion exchange membrane into the acid stream, where they outgas as CO2 and are collected [17].
Point-source capture methods with aqueous solvents typically use dissolved amines, such as monoethanolamine, diethanolamine, or piperazine (Bui et al., 2018). Dissolved amines speed absorption kinetics because they react quickly with CO2.6 Consequently, amine solutions suffer from a similar problem to that of strong bases: they have a high CO2 absorption energy and therefore high regeneration energy, albeit slightly less regeneration energy than that of strong bases. Amine-bearing chemicals are not used in solvent-based DAC because their volatility results in high CapEx for preventing evaporative losses when used with the large air contactors that are required for DAC (Kiani et al., 2020; Erans et al., 2022).
1.5 Degassing ocean water in DOC
Ocean water has a DIC concentration of only ~2mM [20], in comparison to ~1 M in a loaded DAC solvent.7 Thus, relative to CO2 regeneration in DAC and point-source capture, DOC faces the challenge of removing CO2 from a ~500-fold larger quantity of aqueous solution using a smaller partial pressure differential. To achieve a large total CO2 removal under those conditions, vacuum pressure (a pressure swing) is used to maximize the partial pressure differential, and high-surface-area hollow fiber membranes are used to compensate for the low outgassing rate (de Lannoy et al., 2018; Digdaya et al., 2020) (Fig. 5B).
The large hollow fiber membrane system accounts for a large component of CapEx. To minimize the quantity of membranes required, DOC methods maximize the outgassing rate by acidifying the ocean water to the point where kinetics are limited by diffusion and membrane resistance rather than bicarbonate dehydration (Chengxiang Xiang, personal communication). The acid serves to increase the rate of CO2 outgassing by increasing the kinetics of dehydration of bicarbonate (Fig. 6A) and by increasing the equilibrium partial pressure of CO2 in the seawater. Acidification is accomplished by acidifying seawater using bipolar membrane electrodialysis before degassing and by correspondingly basifying the DIC-depleted seawater before release into the ocean. The company Captura offers an example of commercial DOC using this approach.
2. Addressable constraints on carbon removal using aqueous solutions
2.1 Cost
Cost is the primary constraint on both DAC and point-source capture. DAC is a particularly attractive form of CDR from the standpoint of durability and environmental impact. However, it is also the most costly: estimates range widely from $100-$1000/tCO2 for capture only (excluding transport and storage, which applies also to the costs described below) (Erans et al., 2022; Young et al., 2022). Because DAC has yet to be implemented at full scale, the cost estimates we discuss throughout are estimated costs. Recent TEAs have provided cost estimates for various DAC, point-source capture, and DOC methods:
Calcium-looping DAC has an estimated cost of $94-$232/tCO2 at scale under favorable scenarios (Keith et al., 2018). It is one of the CDR methods with highest technology readiness. The primary cost factors are the CapEx for the air contactor and the OpEx for energy used for solvent regeneration.
DAC based on an electrochemical pH swing driven by bipolar membrane electrodialysis has an estimated cost of $773/tCO2, with room for improvement through ongoing technology development (Sabatino et al., 2020). The primary cost factors are OpEx for energy used to generate the pH swing and for replacing ion exchange membranes, as well as CapEx for the air contactor.
The estimated cost for CO2 capture after biomass combustion is ~36-53 $/tCO2. That cost is a primary cost component for BECCS, on par with biomass production from forestry residues and agricultural waste, which are estimated at $66/t and $88/t dry biomass (~$36/tCO2 and $48/tCO2), respectively (NASEM, 2019). Although CO2 capture by BECCS is cheaper than DAC, its scalability is limited by biomass availability, both in total and due to competition for biomass with other industries.
Estimated costs for current electrochemical DOC designs range widely from ~$500-2000/tCO2. Lower costs are achieved through integration with the water-pumping systems of desalination plants or power plants using seawater cooling (de Lannoy et al., 2018; Digdaya et al., 2020). The electrodialyzer and energy for electrodialysis are leading cost factors for electrochemical DOC.
2.1.1 Energy for solvent regeneration
For both DAC and point-source capture, the energy required for solvent regeneration is both the primary operating cost constraint and a primary overall cost constraint. The minimum energy requirement for separating CO2 from air is ~0.43 GJ/tCO2(Keith et al., 2010), but the energy usage of methods under development is estimated to be 1-10 GJ/tCO2(Erans et al., 2022). DAC and point-source capture methods in development include calcium looping and electrochemical pH swing:
Carbon Engineering’s calcium looping method requires an estimated 4 GJ/tCO2 for solvent regeneration (Keith et al., 2018).
Regeneration energy estimates for bipolar membrane electrodialysis-driven electrochemical pH swing methods range from 5.4 GJ/tCO2 for an energy-optimized process to 21.7 GJ/tCO2 for a cost-optimized process based on current technology (Sabatino et al., 2020). Electrochemical pH swing methods that rely on aqueous redox carrier molecules such as water-soluble quinones can theoretically achieve lower energies (e.g., an estimated lab-scale energy of 1.1 GJ/tCO2(Jin et al., 2020)), but they incur very high materials costs (Sharifian et al., 2021). However, such aqueous redox methods are at very early technology readiness levels.
Additionally, methods using thermal regeneration face energy cost constraints arising from the temperature requirements in addition to total energy requirements:
Calcium looping requires heating to 900 ˚C, which is too hot to use waste heat and too costly for electric heating,8 thus requiring the burning of fossil fuels based on current energy prices (Keith et al., 2018).
For the leading point-source capture methods using a monoethanolamine (MEA) solution as a solvent, the thermal energy required for solvent regeneration is estimated at ~3-4 GJ/t CO2(Gilassi et al., 2020).
For DOC, electricity for electrodialysis accounts for an estimated 15-20% of total cost [24]. Captura’s underlying technology [25] is estimated to require 3.5 GJ/t CO2 for the electrodialysis, and a recently-developed electrochemical pH swing is estimated to require 2.8 GJ/t CO2[13].
2.1.2 CapEx of air contactors, calciners, and absorption towers
A leading CapEx cost in aqueous solvent DAC is the air contactor, which must be very large in order to sufficiently contact air for absorption of CO2:
For calcium looping, the air contactor accounts for an estimated ~32% of CapEx and ~18% of total cost (NASEM, 2019).
For bipolar membrane electrodialysis-driven electrochemical pH swing, the air contactor is estimated at 58% of CapEx and 8% of total cost [16].
Calcium looping requires a calciner to extract gaseous CO2 from solid calcium carbonate pellets as part of the overall process of sorbent regeneration. The calciner is also a leading CapEx cost, accounting for an estimated ~41% of CapEx and 23% of total cost (NASEM, 2019).
For point-source capture, the absorption tower is a primary contributor to CapEx.
2.1.3 Cost of membranes and electrodialyzer for electrodialysis
For electrochemical DAC, the bipolar membranes used for separation of acid, base, and (bi)carbonate ions, are a leading cost constraint. For electrochemical pH swing driven by bipolar membrane electrodialysis, membrane replacement accounts for 37% of OpEx and 32% of total cost (Sabatino et al., 2020). The required membrane area, and therefore the cost, is proportional to the current efficiency, which is the ratio between the number of moles of carbon transported across the electrochemical membrane and the charge transported.
For DOC, the combined CapEx and OpEx for using electrodialysis has been estimated at ~30-40% of total cost, with CapEx of the electrodialyzer accounting for ~14-33% of that and electricity accounting for nearly 50% of it (Eisaman et al., 2018). [25] estimated the cost of the electrodialyzer more conservatively at ~2% of total cost.
2.2 Scaling
2.2.1 Regeneration energy
For both DAC, point-source capture, and DOC, the energy required for CO2 regeneration is a scaling constraint, particularly in a scenario where low-emissions energy is limited. For example, given that total world energy consumption was 6 x 1011 GJ in 2021 (Ritchie, 2022), the 4 GJ/tCO2 required for solvent regeneration in calcium-looping DAC indicated by (Keith et al., 2018) would require 6.5% of today's world energy usage to remove 10 Gt/year of CDR.9 As discussed above, high regeneration temperature (e.g., 900 ˚C in calcium looping with thermal regeneration) is also a cost and energy constraint: because it cannot easily leverage renewable electricity (due to cost) or waste heat, it typically requires burning fossil fuels.
3. Possible solutions
3.1 Carbonic anhydrase as a catalyst in direct air capture
Figure 7. Carbonic anhydrase (CA) enhances CO2 absorption by catalyzing CO2 hydration, which is the rate-limiting step in CO2 absorption in DAC and point-source capture. Catalysis of CO2 absorption via immobilized or free CA may improve cost and scaling of aqueous solvent-based DAC and point-source capture by 1) enabling the use of solvents with lower regenerations energies, thus reducing energy costs, and/or 2) reducing the required size of the air contactor, thus reducing CapEx.
3.1.1 Concept
Recall that CO2 hydration is the rate-limiting step in CO2 absorption for most CDR systems using aqueous solvents. Because carbonic anhydrase (CA) catalyzes CO2 (de)hydration extremely efficiently (Note 1A: Carbonic anhydrases), it has significant potential for use as a catalyst in DAC to enable fast CO2 absorption kinetics in aqueous solvents. Catalyzing CO2 absorption would enable use of solvents with small enthalpies of absorption that have slow absorption kinetics in the absence of a catalyst, such as weak bases (e.g., potassium carbonate (K2CO3)), or further enhance absorption rate in strong bases. That arrangement would resolve the trade-off between absorption kinetics and regeneration energy (Table 1), and would in turn reduce the energy required for solvent regeneration and/or enable the use of smaller air contactors or smaller electrochemical membranes (Fig. 7, Table 2).
Additionally, absorbing CO2 at the milder pH ranges afforded by weak bases could achieve up to a two-fold increase in energetic efficiency due to increased stoichiometric efficiency of alkalinity consumption during the absorption process [17], on top of reduced regeneration energy. The stoichiometric efficiency stems from CO2 being absorbed as predominantly bicarbonate (HCO3-) below its pKa2 of pH 10.3, which consumes only one unit of alkalinity from the solvent, rather than as carbonate (CO32-), which predominates at high pH and consumes two units of alkalinity (Fig. 8A). Using CA to catalyze dehydration during regeneration could further reduce regeneration energy, allowing faster outgassing under milder conditions. Distinctly optimized CAs could be used for absorption and solvent regeneration, if the CA were immobilized and thus not required to pass through both processes.
Figure 8. A) Units of alkalinity consumed per unit CO2 as a function of solvent pH. At low pH, CO2 predominantly takes the form of HCO3-, which consumes one unit of alkalinity via Fig. 6A,Eqn 1 or 2. At high pH, CO2 predominantly takes the form of CO32-, which consumes an additional unit of alkalinity via Fig. 6A,Eqn 4. B) Normalized energy requirements for electrochemically separating OH- and H+ to produce acid and base for pH swing, as a function of ∆pH (size of pH swing). Note that this calculation ignores overpotential.
Although use of CA in DAC has been explored sparingly, the promise of CA is well acknowledged with respect to point-source capture: substantial related research is ongoing (Salmon & House, 2015; Bhagat et al., 2017; Effendi & Ng, 2019; Molina-Fernández & Luis, 2021; de Oliveira Maciel et al., 2022); over 100 patents were registered between 2016 and 2022 (de Oliveira Maciel et al., 2022); and Sapiem has planned commercialization in collaboration with enzyme manufacturer Novozymes (de Oliveira Maciel et al., 2022). Results in DAC results similar to those from point-source capture would offer substantial benefits.
As we describe in depth in Section 3.1.3 (Challenges) below, applying CA in aqueous solvent DAC will involve the challenges of maintaining CA activity and stability under process conditions, positioning a high concentration of CA near the air-solvent interface, immobilizing CA within composite materials, and avoiding biofouling. We propose that these issues can be addressed using protein engineering, metagenomic discovery, and materials engineering, in combination with process engineering.
The potential benefits of using CA to catalyze CO2 absorption and outgassing are well illustrated by the example of DAC based on a pH swing using bipolar membrane electrodialysis [17]. TEA of that method found that the optimal process operated at a pH swing from pH ~2.5 to pH ~13, had a final pH of 11.7 in the loaded solvent, consumed 957 kJ/mol CO2 (21.7 GJ/tCO2), and cost $773/tCO2. The authors commented that this infeasible energy requirement and cost resulted from an optimality tradeoff between capture and regeneration (Sabatino et al., 2020).
Reducing energy requirements and cost could be accomplished in several ways if a weak base solvent with a small pH swing could be used while maintaining high CO2 absorption rate via CA catalysis. First, the energetic cost of dissociating water into acid and base, which is the primary energetic cost, would be reduced by operating over a small pH differential, because the energy required to separate OH- and H+ is proportional to the pH difference of the two solutions (Fig. 8B) [31]. Second, the current efficiency—CO2 captured per unit charge transported—would be doubled due to CO2 being absorbed as bicarbonate rather than carbonate, which consumes only one OH- per CO2 (Fig. 8A). For a quantitative comparison, consider a CA-catalyzed process with a smaller pH swing that uses a capture solvent at pH 9, a loaded solvent at pH 8, and an acid stream at pH 6, with CA leveraged for both capture and outgassing (the equilibrium ratio of bicarbonate/CO2 varies ~100-fold between pH 6 and 8, (Fig. 6A)(Datta et al., 2013)). The smaller swing would require 71% less energy to separate OH- and H+ ions and would also use the ions twice as efficiently, resulting in an overall 86% reduction of the energy required for water dissociation per unit of CO2 captured (3.0 GJ/tCO2, down from 21.7 GJ/tCO2).10
The cost reduction could also come in the form of reduced membrane area, which scales inversely with current efficiency (Sabatino et al., 2020), and/or through reduced air contactor size, which scales inversely with the kinetics of CO2 absorption (Fig. 4). Energy and membrane replacement are the primary OpEx costs for DAC using bipolar membrane electrodialysis, and the air contactor and membranes are the primary CapEx costs, so the overall cost reduction considered here would be substantial (Sabatino et al., 2022).
Thermal swing methods offer similar advantages. The calcium looping method described by (Keith et al., 2018) uses a KOH solvent at pH ~14 to absorb CO2 as carbonate. The resulting loaded solvent is only slightly less basic than the initial capture solvent (still near pH 14), due to optimization of absorption rate at the expense of low solvent loading [16]. The carbonates are then precipitated as solid calcium carbonates and calcined to release the CO2, even though regenerating CO2 from solid carbonates requires more energy per mole than regeneration from dissolved carbonates. Solid carbonates are chosen because the temperature required to regenerate dissolved carbonates from a basic solvent is so high that heating of the water component of the solvent renders thermal regeneration from the aqueous phase inefficient, thus favoring heating of solid carbonates to an even higher temperature in the absence of water. However, if a high CO2 absorption rate could be obtained via CA catalysis, a weak base solvent could be used, which would have a lower enthalpy of absorption and would require a substantially lower temperature to outgas CO2.
In a study experimenting with solvents for point-source capture, a 20% potassium carbonate solvent containing CA was shown to afford regeneration at low temperature (50-70 ˚C) under a vacuum (Lu et al., 2011). Such a low regeneration temperature would enable substantial energy reduction, and allow the use of waste heat and/or renewable electricity for regeneration, as is done with many solid sorbents and the amine-based aqueous solvents used in point-source capture. Importantly, it would also avoid the need for precipitating carbonate into a solid phase, obviating multiple process components (the pellet reactor, slaker, and calciner). Taken together, those components are estimated to account for 59% of CapEx, or 33% of overall cost (NASEM, 2019.). While an extra process component would have to be added for regenerating CO2 from the aqueous phase, the net CapEx savings would likely be substantial. Fast absorption kinetics could also reduce the required size of the air contactor, which would afford further savings on CapEx.
Surprisingly, the potential application of CA or other hydration catalysts to DAC is, to our knowledge, almost completely unexplored in publicly-available literature. Several review and overview articles mention it as a possibility (Sclarsic, 2021, Wilcox, 2012; McQueen et al., 2021; Sharifian et al., 2021; Erans et al., 2022). Data in a patent by Carbon Engineering showed a ~20-fold increase in the mass transfer coefficient of 0.5 M K2CO3 (pH 11) along with a more modest increase for other solvent solutions with higher enthalpies of absorption (Henderson, 2014). However, detailed descriptions of potential advantages, exploration of possible designs, optimizations, and ways to circumvent engineering challenges were not given. On the other hand, the potential value of CA in point-source capture is well-acknowledged, as discussed above. In point-source capture, many instances of 5- to 10-fold, and up to 30-fold, improvements in CO2 mass transfer have been obtained under various bench-scale process conditions through enhancement by CA (Alvizo et al., 2014; Fradette et al., 2019; de Oliveira Maciel et al., 2022).11 Absorption tower height was estimated to be reduced by over 95% when CA was used with a sodium carbonate solvent, in comparison to the unfeasibly large tower height required for the solvent in absence of the catalyst Penders-van Elk et al., 2013.
An alternative catalyst for DAC is a CA mimic, a non-protein molecule that mimics features of the CA active site and may have, in principle, many of the same advantages of CA. We focus on CA in this roadmap because it has higher activity and is more engineerable because highly multiplexed production of variants is straightforward via DNA synthesis. However, CA mimics may have several advantages in principle (Note 1B: CA mimics).
From our interviews and literature review, it appears that research into the use of CA in DAC has been hindered for several reasons. First, there is a conception among DAC technologists that CA cannot be durable enough under process conditions. Second, there is a solvable constraint on the cost of CA. Finally, there has been a lack of overlap between DAC and bioengineering researchers, such that the required R&D on enzymatic catalysis is not a favored route among teams.
Constraint addressed
Mechanism
Possible benefit
High solvent regeneration energy (cost and scaling)
Enabling use of weakly basic solvents (e.g. K2CO3) with low regeneration energy but slow absorption without catalysis
Estimate requires process modeling and TEA. For an energy-intensive (21.7 GJ/tCO2) electrochemical DAC method [40], back-of-envelope suggests >80% energy reduction may be feasible
Enabling use of weakly basic solvents (e.g. K2CO3) that have regeneration temperatures low enough to use industrial waste heat or electric heating, but have slow absorption without catalysis
Estimate requires process modeling and TEA.
High CapEx of air contactor (cost)
Enabling smaller contactor via increased rate of absorption of CO2 per unit area
Estimate requires process modeling and TEA. Benefit may be substantial, but may trade off with use of solvents with low regeneration energy.
Contactor area scales inversely with CO2 absorption rate.
High CapEx of calciner for temperature swing DAC (cost)
Obviating calcining in favor of regeneration from liquid phase, by enabling use of weakly basic solvents with low regeneration energy
Obviate ~41% of CapEx, and 23% of total cost (NASEM, 2019), while adding costs associated with regeneration of liquid solvent
High OpEx of bipolar membrane replacement for pH swing DAC (cost)
Increasing current efficiency by absorbing carbon as HCO3- vs CO32-, which inversely determines the required membrane area
Estimate requires process modeling and TEA; back-of-envelop says up to a doubling in current efficiency, or 50% reduction in membrane usage (up to 18.5% and 16% reduction in OpEx and total cost [40])
Table 2. Summary of mechanisms by which CA can address constraints on DAC, and the possible benefits.
3.1.2 Constraints addressed
3.1.2.1 Cost
There is potential to substantially reduce the energy required for solvent regeneration, which is the primary operating cost constraint in aqueous solvent DAC. For example, in the case of pH swing using bipolar membrane electrodialysis, the energy required for electrochemical separation of acid and base is the primary contribution to the energy demand of 21.7 GJ/tCO2 (estimated using current methods). It may be possible to reduce that energy requirement by 86%.
There is potential to reduce the temperature required for solvent regeneration in DAC based on temperature swing, to an extent that enables economical electric heating and/or enables the use of waste heat instead of combustion of fossil fuels.
There is potential to substantially reduce CapEx of aqueous solvent DAC by reducing the size of the air contactor, thereby obviating the calciner and slaker in DAC based on temperature swing, which account for ~41% of CapEx, and 23% of total cost (NASEM, 2019.). (However, other costs would be added through the liquid solvent regeneration process.)
There is potential for up to 50% reduction in the surface area of membranes required for DAC using electrochemical pH swing, or up to 18.5% of OpEx, and 16% of total cost (Sabatino et al., 2020).
3.1.2.2 Scaling
The reductions in energy usage described above would enable scaling with a reduced burden on the supply of low-carbon energy.
3.1.3 Challenges
Table 3 summarizes the challenges of using CA in DAC and gives recommendations for addressing those challenges. Subsequent subsections discuss the challenges in detail, and Section 4 (Recommendations) at the end of this document provides a more detailed list of recommendations.
Challenge
Recommendation
Lack of process designs and TEA integrating CA and DAC to quantify benefits and guide research
Perform combined process modeling and TEA on prospective designs, where CA properties are treated as mutable, to guide research
CA must be near gas-solvent interface
If CA is to be immobilized, develop surface-proximal formats
CA immobilization in effective format
Develop composite materials containing (possibly engineered) CA, e.g. textiles, particles, or membranes, guided by process and cost optimization
Develop platforms for screening performance of immobilized CA under simulated process conditions
Achieving optimal stability, activity, and physicochemical properties of CA in deployment format
Process modeling to indicate target properties
Develop platforms for screening CA in high throughput and/or under simulated process conditions
Metagenomic discovery and protein engineering toward target properties
Develop immobilization formats that enhance enzyme stability, solubility, and/or activity
Biofouling by microbes that feed on CA
Study to predict impact of biofouling expected for process designs, so it can be accounted for in tech dev
Consider impact of biofouling mitigation on process efficiency and cost early in tech dev
Immobilize CA to restrict the presence of biofouling
CA cost
TEAs should anticipate strong cost reduction for CA via economy of scale intrinsic to biomanufacturing when CDR is scaled
Market shaping to increase demand pull, e.g. volume guarantees, could decrease cost during scale up of CDR
Immobilization allows for less CA, with added cost of fabricating composite materials
Constructing interdisciplinary teams
Support formation of teams including protein/materials/process engineering through institutional and funding support
Protein engineering could benefit from prize competitions if target properties clearly defined
Table 3. Summary of challenges to using CA in DAC and recommendations for addressing them.
3.1.3.1 Process designs and TEA integrating CA and aqueous sorbent DAC
Successfully applying CA in DAC will require co-optimization of process conditions, deployment format, and enzyme characteristics. For example, process component designs should be optimized for improved kinetics resulting from catalysis; the CA should be positioned such that its accessibility to CO2 is optimal (see below); and the CA should function well under the deployment conditions (e.g., pH, temperature, immobilization) (see below). Optimizing process designs will specify the desired operating environment of the enzyme. Desired enzyme operating environments will, in turn, determine the target characteristics of the protein and guide efforts to optimize CA using protein engineering and materials engineering. CA could be distinctly optimized for absorption and regeneration, if it is used in an immobilized format, which would reduce the need for trade-offs in the properties of CA used in the respective processes.
We are not aware of published work on process or component designs for aqueous solvent DAC that are optimized for the use of catalysts. That is particularly surprising given the substantial work that has been done on the potential for CA to improve point-source capture (Salmon & House, 2015; Bhagat, Dudhagara, & Tank, 2017; de Oliveira Maciel., 2022). In the context of point-source capture, TEAs indicate the potential for substantial cost benefits through CA use (Reardon, 2015, Gilassi, 2020, 2021). Therefore, TEAs to optimize and evaluate DAC processes leveraging CA should also be performed. Crucially, and in contrast with the TEAs cited, some modeling efforts should also consider enzyme characteristics and deployment format to be mutable (i.e., engineerable), meaning that the enzyme would be treated as a component that could be varied during full-system optimization. Such studies would not only demonstrate the potential benefit of an optimized enzyme, but also identify optimal pH, temperature, alkalinity, and immobilization conditions to provide targets for protein engineers and materials engineers.
3.1.3.2 CA proximity to the gas-solvent interface
The rate of mass transfer of CO2 from gas into aqueous solution depends on disequilibrium between the gas and solvent at the gas-solvent interface.12 Thus, rapid clearing of CO2(aq) from the gas-solvent interface must be achieved for the CO2 absorption rate to be increased. CO2(aq) can be cleared from the interface either by reaction (i.e., hydration) or diffusion, but for practical purposes the hydration reaction must dominate because diffusion is too slow (Fig. 9A) [36]. That is the reason why hydration kinetics are so important for the rate of CO2 absorption (see Note 1C: Hydration kinetics at the air-solvent interface).
Figure 9.A) When CO2 diffuses across the gas-solvent interface, it can be cleared by diffusing into the bulk, which is a slow process, or by reacting to form bicarbonate which is fast at high pH or when catalyzed by CA. B) When CA is positioned close to the gas-liquid interface, clearing of CO2 from the interface can be dominated by reaction kinetics vs. diffusion more quickly, resulting in a lower concentration of CO2 at the interface (and therefore faster absorption). CO2(aq)i, CO2(aq) at the interface; CO2(aq)b, CO2(aq) in the bulk; δCA, distance of CA from interface.
Given the importance of hydration near the interface, it is crucial that CA acts near the gas-solvent interface when it is used to boost hydration kinetics. Otherwise, diffusion of CO2 through the bulk solvent, before it reaches the enzyme, will dominate the mass transfer kinetics, and the enzyme will play only a minor role (Fig. 9B) (Penders-van Elk, et al. 2013; Molina-Fernández & Luis, 2021).
CA is naturally present at the interface when deployed as a dissolved enzyme. However, immobilized CA, which is desirable for several reasons (see Section 3.1.3.3, below), is more difficult to position near the interface. Early efforts that immobilized CA on a solid packing material or porous solid medium, such as polyurethane foam, found that the kinetic enhancement was minor, and substantially lower than free enzyme, unless flow rate was substantially increased [29]. (Panders-van Elk, 2013) studied the kinetic enhancement of CO2 mass transfer produced by CA immobilized on nylon beads of varying sizes, and found that beads <20 µm in diameter improved kinetics up to 50% as well as free enzyme, while larger beads were closer to 10%. They ascribed this effect to a majority of the surface of larger particles not immersing close enough to the gas–solvent interface. Using film theory, they calculated that for a "typical" CO2 absorber (we assume they mean using MEA), the hydration reaction occurs within the top 1-10 µm. Thus, any efforts to deploy CA in DAC should strive to position the enzyme within the first few tens of micrometers, and ideally the first ten.
One potential method for localizing CA to the air–solvent interface is tethering it to a surfactant, thus causing it to associate with the air–solvent interface. It may be possible to engineer surfactant peptides into the CA itself. These peptides can be engineered to lose surfactant properties in response to process parameters such as pH, which could be helpful for reducing foaming (Dexter, Malcolm, & Middelberg, 2006; Dexter & Middelberg, 2008). The achievable effective interface concentration using surface-associated CA may be limited by enzyme size and viscosity (Roger Aines, personal communication) or by enzyme solubility (see Section 3.1.3.3 (CA immobilization), below). Some methods for immobilizing CA can also achieve effective surface association (see Section 3.1.3.3 (CA immobilization), below).
3.1.3.3 CA immobilization
Table 4. Advantages and disadvantages of immobilizing CA in DAC or point-source capture.
Optimizing the format for deploying CA in an aqueous solvent is an active area of research. Initial designs focused on the use of free dissolved enzyme, which has the advantages of being simple and directly accessing the gas-solvent interface. However, using free enzyme also has several problems. First, it requires a large total quantity of CA in the solvent, much of which is not optimally utilized due to its distance from the gas-solvent interface. It also requires that the CA be exposed to multiple process conditions, which can cause faster degradation (for example, due to solvent regeneration at elevated temperatures) and adds complexity to the target stability and activity characteristics for protein engineering. Additionally, it does not allow for the use of distinctly optimized CA for absorption and solvent regeneration. Finally, by exposing the entire process to dissolved protein, it may exacerbate biofouling.
Optimizing the format for deploying CA in an aqueous solvent is an active area of research. Initial designs focused on the use of free dissolved enzyme, which has the advantages of being simple and directly accessing the gas-solvent interface. However, using free enzyme also has several problems. First, it requires a large total quantity of CA in the solvent, much of which is not optimally utilized due to its distance from the gas-solvent interface. It also requires that the CA be exposed to multiple process conditions, which can cause faster degradation (for example, due to solvent regeneration at elevated temperatures) and adds complexity to the target stability and activity characteristics for protein engineering. Additionally, it does not allow for the use of distinctly optimized CA for absorption and solvent regeneration. Finally, by exposing the entire process to dissolved protein, it may exacerbate biofouling.
To circumvent those issues, recent efforts in point-source capture focus on immobilizing the enzyme) to form composite materials used in the absorber (Molina-Fernández & Luis, 2021; Rasouli et al. 2022; Shen, Yuan and Salmon, 2022). Enzyme may be immobilized on materials such as textiles, porous solids, or particles. Immobilization to a textile or membrane maintains the enzyme in a single location. In contrast, immobilization on a bead or particle allows recovery of the enzyme via density gradient, size-exclusion, or magnetic separation before solvent regeneration [29]. Enzyme immobilization can have multiple distinct advantages (Table 4): it can stabilize the enzyme physicochemically, high-density immobilization can increase the effective concentration of the enzyme beyond solubility limits, and it can reduce CA cost by using it only where it is necessary. Immobilization may also reduce the risk of biofouling by keeping the CA in one location rather than throughout all processes, and possibly by physically blocking microbes from accessing the CA. Immobilized CA could also be distinctly optimized for the absorption and regeneration processes, improving performance.
Immobilization methods and substrates have varied widely in recent research. Categories include covalent attachment, such as by chemically reacting exposed moieties on the protein with surfaces, beads or nanoparticles; entrapment, such as within porous alginate beads; and adsorption, including onto polymer beads or metal nanoparticles [29].
Despite the potential benefits, current methods for immobilizing CA do not match the enhancement of CO2 absorption of free CA due to a few challenges that depend on the immobilization method. Most importantly, kinetics can be reduced due to slow diffusion of CO2 through a surface film or through the immobilization substrate itself (e.g., entrapped enzyme) before reaching the enzyme, and the enzyme can be lost due to leaching during operation (e.g., from adsorbed enzyme). Chemical damage to the protein during immobilization can also reduce activity (e.g., covalent attachment) [51].
Several immobilization methods that can maintain CA close to the air-solvent interface to maximize accessibility to CO2 are being developed. These methods, discussed in subsequent sections, include immobilizing CA on high-wetting textiles, buoyant particles, and membranes (Table 5).
(Zhang et al., 2022), (Yong et al., 2016), [57], (Rasouli et al. 2022), (Russo, 2022)
Metal organic framework
[53], (Zhang et al., 2022), (Rasouli et al. 2022)
Table 5. Formats for immobilization of CA.
3.1.3.3.1 Textile immobilization
Immobilizing CA to high-wetting, high-surface area textile materials is a recent approach that is exciting due to its simplicity. Wicking of solvent along the textile forms a very thin surface film, reducing the diffusion limitations that typically make immobilization on packing materials unfavorable (Yuan et al., 2021; Shen, Yuan, & Salmon, 2022). In a recent point-source capture study, CA was immobilized on ~30 µm-wide cotton (cellulose, essentially) fibers by covalent attachment to, or entrapment within, a thin chitosan matrix deposited on the surface of the fibers [52]. In a potassium carbonate solvent, the high surface area of cotton textiles alone—without enzyme—improved CO2 capture rate by ~7-fold in comparison to traditional Raschig ring-structured packing.13 Covalent attachment of CA enabled another 3-fold improvement. Robust CA immobilization resulted from covalent attachment via glutaraldehyde crosslinking, as demonstrated by ~100% activity retention after 500 h of continuous solvent flow. The authors suggested that the capture performance of CA-coated textiles could be improved through optimization of enzyme loading, robust covalent attachment mechanisms that preserve enzyme activity, and attaching or loading CA-immobilized particles on the fibers. As we discuss next, biotechnology can address the latter two strategies.
Attaching CA-immobilized particles to the textile could further increase surface area. CA can be covalently attached to nanoparticles or microparticles [29], and particle-like aggregates of CA can be formed by glutaraldehyde cross-linking (Cui & Jia, 2013; Molina-Fernández & Luis, 2021; Ren et al., 2021). Biologically-assembled particles that are densely coated with CA could also be produced, possibly for lower cost by avoiding fabrication and purification processes. Examples include display of carbonic anhydrase on the surface of diatoms through fusion to the silica-condensing protein R5, which results in covalent attachment of the protein to the siliceous shell of the diatom, obviating the need for costly protein purification (Sheppard, et al. 2012). A similar method using the same fusion proteins can be used to make silica nanoparticles (Hoon Jo et al., 2014). Both methods have very low enzyme leaching and can retain high catalytic activity. Using emerging protein-engineering methods, self-assembling protein nanoparticles bearing CA could also be constructed, either as modified viral capsids or using de novo design [63]. In any of those examples, the particles could be attached to the textile by chemical cross-linking (e.g., glutaraldehyde, as used to anchor CA by [52].
There are several options for enabling robust anchoring and retained activity of cross-linked CA on a textile, particle, or other substrate. Most simply, the chemical identity and concentration of crosslinker can be optimized. Alternatively, protein engineering could be used. For example, it may be possible to associate CA with a textile by fusing CA to a protein domain that binds the textile and then treating the sample with a fixative to crosslink the CA to the textile in an orientation that maximizes structural preservation of the CA. As one possibility, a cellulose-binding domain could bind the cotton textile (Levy & Shoseyov, 2002; Roberts et al., 2021) or a chitosan coating on the textile (Hou et al., 2019; Shen, Yuan, & Salmon, 2022). To increase CA density, multiple CAs could be associated with a single cellulose-binding domain using displayed epitopes and affinity domains, similar to signal amplification using fluorescent secondary antibodies for microscopy. Binding domains to the chitosan matrix coating, or another coating, could also be explored.
Alternatively, using site-specific attachment chemistry could avoid chemical damage to the protein . Site-specific protein attachment is a very active domain of biotechnology research, mostly for applications in biologic pharmaceuticals or biological imaging, and there are many available and emerging chemistries that could be useful for immobilization of CA [67]. For example, a peptide tag or non-standard amino acid that reacts specifically with a chemical or peptide incorporated into the immobilization substrate could be added to CA.
3.1.3.3.2 Particle immobilization
Buoyant particles bearing immobilized CA can also be used to associate CA with the gas-solvent interface. The buoyant particles are then recovered via floatation tank for reuse before solvent regeneration, thus avoiding thermal treatment. CA entrapped in buoyant polysilicate–polysilicone microparticles (Bucholz et al., 2016) conferred a 6-fold increase in CO2 absorption in an amine solvent [68]. Separately, a TEA for a point-source capture system using buoyant CA-immobilized beads in a proprietary salt solution achieved 31.5% cost reduction compared to a 30% MEA solvent (Reardon et al., 2015).
In some formats, particle-immobilization can increase the effective concentration of CA beyond the solubility limit [29]. Particle-immobilized CA can also improve robustness of the enzyme to mechanical stresses introduced for process intensification (Verma et al., 2016; Molina-Fernández & Luis, 2021). It is also possible that particle-immobilized CA may exhibit reduced viscosity per unit activity, because immobilization results in less of the protein surface being available for intermolecular interactions. As mentioned above, it may be possible to use covalently-attached surfactants or surfactant peptides to enhance the surface-association of a diversity of CA-coated particles [29].
3.1.3.3.3 Membrane immobilization
Another strategy for localizing CA to the air-solvent interface is embedding it in membranes (Zhang et al., 2022). Hollow fiber membrane contactors have exceptionally high surface area, and they may be an attractive option for DAC if membrane fabrication prices and membrane resistance to mass transport can be reduced (Diederichsen et al., 2022). For example, adsorbing CA onto porous polypropylene hollow fiber membranes increased CO2 absorption into potassium carbonate by three-fold (Yong et al., 2016). Similarly, Carbozyme, Inc., used CA-embedded polypropylene hollow fiber membranes [70].
Existing membranes generally exhibit a sharp trade-off between CO2 permeance and CO2/N2 selectivity (Wang et al., 2016). Encouraging work in novel, ultra-thin membranes may produce breakthroughs if costs can be reduced (Molina-Fernández & Luis, 2021; Zhang et al., 2022). A particularly exciting example is an ultra-thin nanoporous membrane that confines CA in a nano-film, increasing the effective concentration of CA 10X beyond its solubility, achieving CO2/N2 selectivity of 788 and permeance of 2600 GPU, and enabling a mass transport regime that is limited by enzyme activity rather than diffusion [72].
Membranes are currently limited by cost, but multi-fold improvements in mass transfer via CA catalysis could contribute to cost reduction by reducing the total surface area of membrane required per unit CO2 captured (Iliuta & Iliuta, 2017; Rasouli, Nguyen, & Iliuta, 2022). If membranes are used in an air contactor, the energy and cost of blowing air over the membranes will require specific attention, because that setup may involve a larger pressure drop than for packing with a lower surface area.
3.1.3.3.4 Metal organic framework immobilization
Finally, it is worth noting that immobilization within metal organic frameworks (MOFs) or nanoflowers can yield remarkable benefits for CA activity and stability, including activity enhancement relative to free enzyme [29]. They may need to be immobilized on another support structure because they are mechanically fragile. Therefore, immobilizing these materials into textiles or membranes is an active research direction (Zhang et al., 2022). These emerging materials may allow substantial increases in activity and stability if fabrication costs can be reduced.
In general, enzyme immobilization introduces the additional cost of fabricating composite materials. In the case of DAC, the air contactor is larger than in point-source capture.This means a larger quantity of composite material would need to be fabricated to use immobilized CA in DAC. Despite fabrication costs, and even with the reduced catalytic efficiency that can result from immobilization, using immobilized CA may still result in reduced overall cost compared to free enzyme. The emerging view is that enzyme immobilization will be adopted for point-source capture due to reduced cost of enzyme replacement [29].
Overall, joint efforts in materials engineering and protein engineering should be used to optimize CA activity in the context of immobilization, ideally with target performance informed by system-level TEAs, as done by (Young, et al. 2023) for solid sorbents. It is, however, important to pay attention to the costs associated with using chemically-modified proteins or with using multiple proteins for increased display density. TEAs should be conducted to get a sense for which format is most appropriate for DAC.
3.1.3.4 Stability, activity, and properties of CA in deployment
The performance and physicochemical properties of CA under process conditions and deployment formats influences the viability of process designs using CA. Enzyme stability influences the cost of enzyme replacement, and activity influences the extent to which catalysis can supplant chemical reactivity to boost CA hydration kinetics. Physicochemical properties such as solubility and viscosity of the enzyme in the solvent influence the concentration in which CA can be deployed as a free enzyme.
3.1.3.4.1 CA stability and activity
Due to their strong influence on cost, CA stability and activity have been the primary focus of research into the use of CA in carbon management (with research historically focused on point-source capture). Factors influencing CA activity and stability include temperature, pH, alkalinity, and immobilization format. For point-source capture applications, efforts to engineer or discover natural variants of CA have yielded enormous improvements in CA stability and activity under high temperature, alkalinity, salt, and pH (Mesbahuddin, Ganesan, & Kalyaanamoorthy, 2021; Talekar et al., 2022). Examples include engineered tolerance of exposure to 107 ˚C in 4.2 M MEA for 1 hr (Alvizo et al., 2014), and natural tolerance to heating to 80 ˚C for 3 hr (Del Prete et al., 2017). Protein engineering and metagenomics (the study of genetic sequences derived from microbial communities) are fast-growing fields of biotechnology; their continued efforts will likely produce further improvements. Encouragingly, the process conditions used in DAC solvents composed of weak bases are likely to be milder than those used in point-source capture because lower regeneration temperatures can be used (e.g., 50-70 ˚C in 20% potassium carbonate under a vacuum (Lu et al., 2011) vs. ~80–130 ˚C for the amine-based sorbents used in point-source capture [37]). Indeed, the more efficiently CA can be used to catalyze CO2 (de)hydration, the milder the process conditions will need to be, leading to increasing returns from higher-performing CAs.
Going forward, there are a range of options for improving CA stability and activity. In the near term, improvements may be achieved using new methods for in silico protein engineering based on machine learning (Yang et al., 2019) and large-scale screening of carbonic anhydrases derived from novel or existing metagenomic sequence datasets. In silico methods may be particularly valuable given that screens for carbonic anhydrase activity currently cannot be performed in ultra-high throughput, since they require collection of optical data over a time-course of seconds to minutes (Kim & Jo, 2022). A valuable advance would be development of improved screening methods that enable ultra-high-throughput screening of CA activity, e.g., using droplets, given that existing brute-force screening methods scale only to ~103 variants per screen (Alvizo et al., 2014).
Hyperstable protein structures have been designed with melting temperature midpoints (Tm) of at least 122 ˚C (Kimura et al., 2020), which is much higher than any engineered or known natural CA. Encouragingly, early work showed that CA-like enzymes can be designed de novo by placing Zn2+-coordinating centers on protein scaffolds (Zastrow et al., 2011; Cangelosi et al., 2014). Those efforts achieved activity 3-4 orders of magnitude lower than the fastest CAs, but still faster than small-molecule CA mimics [84]. Emerging methods applied to the same challenge may reasonably be expected to perform better, perhaps matching the activity of natural CAs but exhibiting vastly greater thermotolerance and engineerability.
CA immobilization may also yield improvements to CA stability and activity (Molina-Fernández & Luis, 2021; Rasouli, Nguyen, & Iliuta, 2022).
3.1.3.4.2 Physicochemical properties of CA
CA solubility may be a challenge when using free (non-immobilized) CA in high-capacity solvents. Ideal CA concentrations for mass transfer are 1-2 g/L or more (Salmon & House, 2015; Zhang & Lu, 2015). Those concentrations are readily achievable in solutions such as 0.5 M potassium carbonate (~7% by weight); but for a higher-capacity solvent such as 20% potassium carbonate, the solubility of bovine CA may be lower. Solubility has been reported as low as ~0.1 g/L (Russo et al., 2016), but much higher concentrations of CA have been used in other experiments with concentrated potassium carbonate (de Oliveira Maciel et al., 2022; Table 6), so it is unclear exactly what the constraints of CA solubility are.
However, because the solubility of different proteins can vary by orders of magnitude, it may be possible to increase the solubility of CA in high salt and alkalinity using protein engineering or computational mining of CAs from metagenomic sequences of halophilic communities (Manyumwa et al., 2021) to accommodate high-capacity solvents. Solubility can also be enhanced using CA-immobilized microparticles, which can increase the effective concentration beyond the solubility limit (Russo et al., 2016).
The viscosity of dissolved CA can also cause challenges, for example, when it is used in high concentration at the gas-solvent interface (see discussion of interface proximity above), facilitated by tethering to a surfactant (Roger Aines, personal communication). Viscosity may be addressable by minimizing intermolecular interactions using protein engineering, by adding salts or other small solutes (Hong, et al. 2018), or by screening different CAs for lower viscosity. Viscosity may also be effectively reduced by reducing the size of the CA protein, which would decrease the mass concentration (and thus the viscosity) of the protein required to achieve a given activity. Size could be reduced by searching for smaller variants; by screening deletion mutants, which has been used to minimize size by 60-70% (Shams et al., 2021); or by designing de novo carbonic anhydrases in small protein frameworks (Zastrow et al., 2011; Cangelosi et al., 2014).
3.1.3.5 Biofouling
Mitigating biofouling, microbial biomass buildup in the solvent or on surface biofilms due to microbes that feed on the CA, will be a key challenge in the use of CA in DAC. In contrast to point-source capture, DAC exposes the solvent directly to air, which contains many microbes. This constraint may limit the translation of existing research on the use of CA in solvents. In the case of free enzyme in the solvent, fouling could be distributed throughout the process equipment. In the case of immobilized enzyme, fouling of the support structure could require frequent replacement, driving up cost. There is a risk of biofouling being ignored in fundamental research on CA-enhanced aqueous solvents for DAC, where processes may not be run for long enough for fouling to cause problems. It is crucial that consideration of biofouling be incorporated into research early-on to determine the constraints it places on designs at scale.
As per traditional industrial antifouling methods, materials choices and surface modification can help prevent fouling, particularly through the use of highly hydrophilic surfaces and coatings, or of biocidal chemicals [90]. Immobilized enzymes such as proteases can be used to prevent biofouling (Bachosz et al., 2022), and could perhaps be immobilized adjacent to CA. It will be important to investigate the impacts of such materials on cost and conflicting interactions with CA function, immobilization mechanisms, and the chemistry and processes of DAC.
One possibility is to isolate the enzyme from microbes in the air by embedding or immobilizing the enzyme inside a porous material or membrane that microbes cannot penetrate. Gas-liquid membrane contactors with CA immobilized (or free-flowing) on the liquid side of the membrane have been explored for point-source capture (Zhang et al., 2022). CA can also be immobilized within a hydrogel in beads (Xu et al., 2021), in a thin layer (Zhang et al., 2010), or in silica gel particles (Min et al., 2016). These approaches may increase resistance to mass transfer of CO2, as discussed above, though membrane contactors can also increase the accessible surface area (Diederichsen et al., 2022) and present a very exciting possibility if transfer resistance can be reduced.
Finally, “mirror carbonic anhydrase,” a carbonic anhydrase constructed of amino acids with the opposite chirality of normal biomolecules, would not be digestible by naturally-occurring organisms and would overcome the biofouling issue [95]. Organisms suitable for manufacturing mirror biomolecules are decades off, because their construction would require de novo synthesis of a full cell and its components. However, they have substantial future potential for preventing fouling with the use of industrial enzymes.
3.1.3.6 CA cost
The cost of CA is frequently cited as a barrier for its use in point-source capture (de Oliveira Maciel et al., 2022). However, in the case of point-source capture, TEAs indicate the potential for substantial cost benefits through the use of free (Gilassi et al., 2020, Gilassi, et al. 2021) or immobilized (Reardon, 2015) thermotolerant CA, and Novozymes and Saipem are collaboratively developing a commercial carbon capture process using CA. (Gilassi et al., 2020) found that CA cost 7.4% of the total OpEx, 4.6% of the total cost, and 1.27X the cost of the rest of the solvent material14 (assuming a cost of ~$480/kg for CA15). Such findings indicate that CA is not dominating cost. Improved designs that better leverage the benefits of CA may further improve the cost efficiency of using CA, and scaling enzyme production via established optimization and scaling methods is expected to lead to cost reduction (Zhang, Sun, & Ma, 2017; Tarafdar et al., 2021) (see Note 1D: Cost of enzyme manufacturing). The cheapest industrial enzymes are currently $10-20/kg due to substantial scale and optimization [98]. Therefore, the cost of CA can likely be reduced by 10- to 20-fold, or more, if it is used in carbon management at scale, which would result in it contributing < 1% of cost in the Gilassi study. This estimate paints an optimistic picture for the use of CA in the case of point-source capture. Given that DAC requires a larger gas-solvent contact area than point-source capture does, cost is expected to be a greater challenge in DAC, and should be explored using TEA.
Cost reduction of CA manufacture could also be hastened by market-shaping methods to enhance demand pull, such as volume guarantees if DAC using CA shows robust promise.
As discussed above, further reduction in the cost of CA may be achievable by applying emerging methods in protein engineering to extend the lifetime of CA under process conditions, reducing replacement rate, and/or by immobilizing CA on a solid support or recoverable particle, reducing the total quantity of CA needed.
3.1.3.7 Constructing interdisciplinary teams
The paucity of work on CA in DAC illustrates that there is a significant lack of teams that combine the required sets of expertise—protein engineering, materials engineering, and process engineering—to optimally address the challenges we have laid out. Avenues to address this lack include establishing large, competitive grants to support interdisciplinary teams; further incorporating bioengineering into carbon management research centers and initiatives; giving increased attention to climate tech in university bioengineering curricula and faculty hires; and community-building at the interface of biotechnology and climate tech. Protein engineering work on CA in DAC may benefit particularly from prize-based competitions if clear target properties can be defined for CA and an assay is developed to evaluate variants relative to the target.
3.2 Carbonic anhydrase or its mimics as a catalyst in direct ocean capture of CO2
Figure10.A) Current techniques use bipolar membrane electrodialysis (pH swing) and hollow fiber membranes under vacuum pressure (pressure swing) to degas CO2 from ocean water. B) Immobilized CA or CA mimics within the hollow fiber membranes catalyze CO2 dehydration, allowing fast degassing in the absence of acidification. That may eliminate the need for a pH swing, or at least reduce the size of the required pH swing.
3.2.1 Concept
As discussed above, leading DOC cost factors are the energy and CapEx of electrodialysis and the electrodialyzer, which are used to acidify the seawater to accelerate outgassing of CO2. Vacuum pressure is also used to cause CO2 to outgas from the seawater. In principle, the vacuum pressure on its own is enough to cause CO2 to outgas; however, the rate of CO2 flux across the membranes is limited by bicarbonate dehydration kinetics16 (Fig. 6A) and by the low transmembrane pressure gradient of CO2,17 both of which are addressed by acidification.
To catalyze bicarbonate dehydration and reduce the quantity of acid needed for outgassing, CA or its mimics could be immobilized on the interior of the hollow fiber membranes used for degassing (Fig. 10, Table 6). In the best case, a bicarbonate dehydration catalyst could enable CO2 to be removed from unaltered seawater. That arrangement would obviate the need for an electrodialyzer (Fig. 10B), thereby removing 30-40% of the estimated cost of DOC (Eisaman et al., 2018) and saving 2.8-3.4 GJ/t CO2 (Digdaya et al., 2020; Kim et al., 2023) while possibly incurring additional energy costs for increased usage of vacuums. If some acidification is still needed in order to boost the equilibrium concentration of CO2 in the seawater, a portion of the energy used for dialysis, which accounts for 15-20% of total cost (Eisaman et al., 2018), could still be saved by using less acidification than current methods (a current method uses pH ~4.7 [99]). This opportunity should be explored first via modeling, and if promising, via experiments.
To our knowledge, existing research on the use of CA and hollow fiber membranes for applications in carbon management has focused on the absorption step in point-source capture rather than outgassing of CO2 (Molina-Fernández & Luis, 2021; Zhang et al., 2022). However, some research has explored CA in hollow fiber membranes for removing CO2 from blood (Arazawa et al., 2012; Arazawa, Kimmel, & Federspiel, 2015).
We note that degassing in DOC may not require the full catalytic capacity of CA because membrane resistance to CO2 diffusion is expected to become rate-limiting. Indeed,(Arazawa et al., 2015) found that membrane resistance was the limiting factor after CA in hollow fiber membranes boosted CO2 removal by 115% from a phosphate buffer and by 36% from blood. Therefore, a CA mimic may prove to be more useful, since it can be more chemically stable even if it is orders of magnitude less efficient as a catalyst (Zhang et al., 2022) (Note 1B: CA mimics).
As discussed in Section 3.2.3 (Challenges) below, the challenges to implementation of CA or its mimics with hollow-fiber membranes for DOC may include the long-term stability of immobilized enzymes, membrane resistance to mass transport, cost of CA, and biofouling of the membranes.
Constraint addressed
Mechanism
Possible benefit
Electrodialysis OpEx
Reducing required size of, or obviating, pH swing by increasing CO2 out-gassing rate at higher pH
Up to the full cost of electrodialysis OpEx (>25% of total cost [102]), on top of CapEx savings, while adding enzyme cost and possibly additional vacuum and membrane costs
Electrodialyzer CapEx
Obviating pH swing by increasing CO2 out-gassing rate at higher pH
Obviating pH swing could save CapEx of electrodialyzer (up to ~10% [102] or ~2% (Digdaya et al., 2020) of total cost), on top of OpEx savings, while adding enzyme cost and possibly additional vacuum and membrane costs
Table 6. Summary of mechanisms by which CA or its mimics can address constraints on DOC and the possible benefits.
3.2.2 Constraints addressed
3.2.2.1 Cost
There is potential for substantial reduction in the cost of DOC associated with electrodialysis, including the cost energy required for electrodialysis and possibly even the CapEx of the electrodialyzer. In a best-case scenario based on cost estimates for DOC, this could amount to 30-40% total cost savings (Eisaman et al., 2018), minus the cost of the catalyst, the process of catalyst immobilization, and any increases in membrane surface area or vacuum pressure that may be needed.
Because CA or CA mimics introduce a new cost (manufacturing and immobilization), they are only beneficial if the savings outweigh their cost.
3.2.3 Challenges
Table 7 summarizes the challenges to using CA in DAC and recommendations for addressing them. We discuss the challenges in detail below, and Section 4 (Recommendations) at the end of this document gives a more detailed list of recommendations.
Challenge
Recommendation
Lack of process designs or TEAs using catalysts in DOC to quantify benefits and guide research
Perform combined process modeling and TEA on prospective designs
Long-term stability of immobilized CA and CA mimics
Process modeling to indicate whether CA would be more or less efficient vs CA mimics
Use of stable CA mimics
Protein engineering to produce hyperstable CA if catalytic efficiency of CA beyond mimics could be helpful
Membrane resistance to mass transport
Development of membranes with embedded catalysts with minimal resistance to mass transport
Cost of CA
TEAs should consider strong cost reduction for CA via economy of scale intrinsic to biomanufacturing when CDR is scaled
Market shaping to increase demand pull, e.g. volume guarantees, could decrease cost during scale up of CDR
Biofouling
Study to predict impact of biofouling expected for process designs, so it can be accounted for in tech dev
Consider impact of biofouling mitigation on process efficiency and cost early in tech dev
Need for interdisciplinary teams
Support formation of teams including materials/process engineering, and protein engineering (if CA is to be used vs a mimic) through institutional and funding support
Table 7. Summary of challenges to using CA in DOC and recommendations for addressing them.
3.2.3.1 Process designs and TEA for integrating catalysts into DOC
It is not clear what outgassing rate is needed from hollow fiber membranes with embedded catalysts in order to achieve cost benefits from only using vacuum pressure and excluding pH swing via electrodialysis, because increased vacuum pressure and larger membrane area might also be required. Combined process modeling and TEA could provide clear engineering targets for membrane performance.
3.2.3.2 Long-term stability of immobilized CA and CA mimics
While improving the rate of CO2 outgassing using catalysts of CO2 dehydration may reduce the membrane area required for DOC, that improvement will only result in a reduction of total membranes if the lifetime of the catalyst-enhanced membranes is not substantially reduced relative to unmodified membranes. Specifically, if the enzyme reduces the required membrane area by a factor of X, then it must reduce the lifetime of the membrane by no more than a factor of X in order to reduce cost.18 Ideally, the membrane replacement rate would not be impacted at all by the use of a catalyst, which would require the catalyst to remain stable throughout the lifetime of the membrane or to be replaced without replacing the membrane itself. Recent cost modeling of DOC assumes replacement of hollow fiber membranes every five years [99].
A promising option for incorporating a long-lived catalyst of CO2 dehydration into a hollow fiber membrane is a CA mimic molecule such as zinc cyclen, rather than CA itself. CA mimics can be incorporated into polymers in membranes, are more stable than proteins at moderate temperatures, and are not vulnerable to proteases that may be present in ocean water (Zhang et al., 2022). While mimics’ catalytic activity is not as high as CA, stability is more important than overall activity in DOC because diffusive resistance will limit the rate of mass transfer of CO2 through hollow fiber membranes after a several-fold boost in dehydration kinetics.
CA engineered for stability is also an option. In addition to having stability with respect to thermal fluctuations, CA would need to be robust against microbial proteases present in ocean water. The best example of stable CA to our knowledge is a variant with a 271-day and 203-day half-life at 40 ˚C and 60 ˚C respectively in 20 mM Tris 300 mM NaCl, pH 8.3 [103]. In the application we are considering for DOC, the CA would be exposed to ambient temperature and a pH at least as moderate as 4.7 [99]. Moreover, emerging enzyme immobilization methods required for application in hollow-fiber membranes can increase stability (Molina-Fernández & Luis, 2021; Rasouli, Nguyen, & Iliuta, 2022), as discussed above. The longest claimed enzyme lifetime we are aware of is an immobilized glucose isomerase, which Novozymes has claimed lasts up to 400 days (Novozymes, no date). To enable extremely long enzyme lifetimes, emerging methods for de novo protein engineering may be used to install a CA active site into de novo designed hyperstable protein structures (Burton et al., 2016; Huang et al., 2016; González‐Castro et al., 2020; Talekar et al., 2022). Nonetheless, a five-year enzyme lifetime is a significant demand beyond existing examples of industrial enzyme usage. Long-term experimentation or access to proprietary data will be needed to determine current benchmarks.
3.2.3.3 Membrane resistance to mass transport
CO2 flux across the hollow fiber membrane is ultimately determined by the transmembrane CO2 pressure gradient of and the membrane resistance to gas diffusion. Vacuum pressure provides low CO2 partial pressure external to the membrane, and dissolved CO2 within the seawater provides the high pressure that drives CO2 across the membrane. At a low enough equilibrium concentration of dissolved CO2, membrane resistance to diffusion will become rate-limiting, at which point catalysis of bicarbonate dehydration will no longer be useful because catalysis can only accelerate equilibration. For example, [107] find that high fluid velocity does not meaningfully accelerate CO2 flux from seawater into an NaOH solvent via a liquid/liquid hollow fiber membrane contactor, because the CO2 concentration in the seawater is low enough that membrane transport is limiting. For such systems, membrane technology innovations that reduce resistance to CO2 transport will unlock greater benefits than bicarbonate dehydration catalysts.
3.2.3.4 CA cost
Enzyme cost is a potential challenge in DOC just as it is in DAC. Recent TEAs indicate that point-source capture using CA is cost-competitive even at current enzyme costs (see Section 3.1.3.6 (CA cost) above), but such evaluations have yet to be done for DAC. Cost reduction of CA due to increases in manufacturing scale are expected to further improve cost-effectiveness. The same methods for cost reduction apply to DOC as discussed above for use of CA in DAC, including the standard practices of scaling biomanufacturing (see Note 1D: Cost of enzyme manufacturing); ongoing efforts to improve biomanufacturing; immobilization of the enzyme to protect it from unnecessary extremes in process conditions and reduce the total quantity needed; market-shaping methods to enhance demand pull, like volume guarantees; and improvements to enzyme stability and performance through protein engineering and materials engineering.
3.2.3.5 Biofouling
Biofouling of the hollow-fiber membranes in DOC may also be a concern. Proposed DOC methods for reducing biofouling involve ultrafiltration of input water to remove bacteria that may colonize membranes [99]. Nevertheless, immobilized proteins or organic molecules like CA mimics may increase the risk of biofouling by cells that do enter the system because they are a food source. If biofouling becomes an issue, immobilized enzymes such as proteases can be used to prevent biofouling [91], and could perhaps be immobilized adjacent to CA, as mentioned above.
3.2.3.6 Constructing interdisciplinary teams
Teams spanning materials engineering, process engineering, and protein engineering (if CA is to be used, rather than a mimic) are needed to most effectively optimize the benefits of catalysts in DOC. Institutional and funding support for cross-disciplinary work would be highly beneficial for developing and sustaining such teams.
3.3 Carbonic anhydrase as a catalyst in point-source capture of CO2
3.3.1 Concept
Point-source capture technology is relevant to CDR in the case of BECCS via biomass combustion, gasification, or pyrolysis.19 It is well-acknowledged that CA (Note 1A: Carbonic anhydrases) has potential to reduce the solvent regeneration energy and absorption tower CapEx required for point-source capture, which are the primary cost and scaling constraints (Fig. 7, Table 8). As we discussed extensively above in the context of DAC, these benefits are the result of catalysis enabling faster CO2 absorption and outgassing kinetics in aqueous solvents. Those faster kinetics allow the use of solvents with lower enthalpies of absorption (e.g., weak bases, like potassium carbonate) compared to the standard amine solutions, and/or allows the use of amine solutions with smaller gas-solvent contact areas while achieving the same CO2 capture rate.20
Research into the use of CA in point-source capture is ongoing. The primary focus to date has been enhancing thermostability, and more recently, deploying CA in an immobilized format, to extend the lifetime of the protein under process conditions (Penders-van Elk et al., 2013; Alvizo et al., 2014; Salmon & House, 2015; Fradette et al., 2017; Bhagat et al., 2017; Effendi & Ng, 2019; Molina-Fernández & Luis, 2021; (de Oliveira Maciel et al., 2022); Rasouli, Nguyen, & Iliuta, 2022). Over 100 patents related to the use of CA in point-source capture were registered between 2016 and 2022 (de Oliveira Maciel et al., 2022), and a capture technology leveraging thermostable CA is presently under commercial development as a collaboration between Novozymes and Saipem (formerly CO2 Solutions, and incorporating advances by Codexis) (Penders-van Elk et al., 2013; Alvizo et al., 2014).
A TEA by Reardon, 2015, estimated that total cost of a point-source capture system was reduced up to 31% when using CA21 immobilized on microparticles in a salt solution solvent (possibly K2CO3), as compared to 30% MEA.22 The key cost reduction driver was a 41% reduction in regeneration energy.23 Protein engineering and materials engineering will enable improvements to the stability, performance, and deployment format of CA that will further improve the benefits of enzyme-catalyzed point-source capture.
[39] predicted CA could enable >90% reduction in absorption tower height when using a potassium carbonate solvent, which may enable the use of towers with comparable sizes to those used with standard amine solutions while achieving lower regeneration energies.
Constraint addressed
Mechanism
Possible benefit
High solvent regeneration energy (cost and scaling)
Enabling use of weakly basic solvents (e.g. K2CO3) with low regeneration energy but slow absorption without catalysis
TEA indicated 41% energy reduction and 31% cost reduction using CA in non-volatile salt solution vs 30% MEA solvent (Reardon 2015), which could be improved via further engineering
Absorption tower CapEx
Enabling smaller absorption tower via increased rate of absorption of CO2 per unit area
Not quantified, may be substantial
Table 8. Summary of mechanisms by which CA can address constraints on point source capture, and the possible benefits.
3.3.2 Constraints addressed
The primary benefits of CA in point-source capture are in terms of cost, and are similar to those for DAC discussed above. These would reduce the cost of BECCS, while also reducing costs across point-source capture in other applications.
3.3.2.1 Cost
There is potential for substantially reducing the energy required for solvent regeneration, which is the primary operating cost constraint in point-source capture and also a primary cost of BECCS. Reardon, 2015, estimates 41% energy reduction and 31% cost reduction, with room for improvement.
There is potential for substantially reducing CapEx by reducing the size of the absorption tower, which is the primary CapEx constraint in point-source capture.
There is potential to use less corrosive solvents (e.g., potassium carbonate rather than MEA), allowing the use of cheaper materials with less frequent replacement.
3.3.3 Challenges
While designs for solvent-based point-source capture of CO2 differ from those for DAC due to different initial concentrations of CO2 in the input gas, the challenges to CA deployment are conceptually similar.
Table 9 summarizes the challenges to using CA in point-source capture, and recommendations for addressing them. The challenges are discussed in detail below, and Section 4 (Recommendations) at the end of this document gives a more detailed list of recommendations.
Challenge
Recommendation
CA must be near gas-solvent interface
If CA is to be immobilized, develop surface-proximal formats
CA immobilization in effective format
Develop composite materials containing (possibly engineered) CA, e.g. textiles, particles, or membranes, guided by process and cost optimization
Develop platforms for screening performance of immobilized CA under simulated process conditions
Achieving optimal stability, activity, and physicochemical properties of CA in deployment format
Process modeling to indicate target properties
Develop platforms for screening CA in high throughput and/or under simulated process conditions
Metagenomic discovery and protein engineering toward target properties
Develop immobilization formats that enhance enzyme stability, solubility, and/or activity
Biofouling by microbes that feed on CA
Study to predict impact of biofouling expected for process designs, so it can be accounted for in tech dev
Consider impact of biofouling mitigation on process efficiency and cost early in tech dev
Immobilize CA to restrict the presence of biofouling
CA cost
TEAs should anticipate strong cost reduction for CA via economy of scale intrinsic to biomanufacturing when CDR is scaled
Market shaping to increase demand pull, e.g. volume guarantees, could decrease cost during scale up of CDR
Immobilization allows for less CA, with added cost of fabricating composite materials
Constructing interdisciplinary teams
Support formation of teams including protein/materials/process engineering through institutional and funding support
Protein engineering could benefit from prize competitions if target properties clearly defined
Table 9. Summary of challenges to using CA in point source capture and recommendations for addressing them.
3.3.3.1 Process designs integrating CA and point-source capture
CA should be deployed in a format where the enzyme and process conditions24 are mutually optimal. We are aware of a few studies that performed an integrated TEA to optimize process design for the use of CA in point-source capture (Reardon, 2015; Gilassi, Taghavi, Rodrigue, & Kaliaguine, 2020, 2021). However, those simulations held CA performance and format static. Future modeling should also vary the enzyme cost and performance (i.e., activity and lifetime) to estimate the benefits of enzyme improvements and provide targets for protein engineers. TEAs should also consider immobilized CA. For optimizing the efficiency of specific process components, the considerations are conceptually similar as for DAC, which are discussed above.
3.3.3.2 Proximity of CA to the gas-solvent interface
As discussed above, proximity to the gas-solvent interface is a key factor in catalyst efficiency (Fig. 9). The methods discussed above for optimizing CA concentration at the interface in the context of DAC also apply to point-source capture. They include immobilizing CA to high-wetting textiles, surface-associating particles, or membranes.
3.3.3.3 Stability, activity, and properties of CA in deployment
CA stability and activity under process conditions such as high temperature, alkalinity, salt, and pH are also key factors in determining the overall cost and efficiency of CA in point-source capture. Advances have been made through protein engineering (Mesbahuddin, Ganesan, & Kalyaanamoorthy, 2021; Talekar, Jo, Dordick, & Kim, 2022) and searching for naturally-stable variants of CA (e.g., engineered tolerance of exposure to 107 ˚C in 4.2 M MEA for 1 hr (Alvizo et al., 2014), and natural tolerance to heating to 80 ˚C for 3 hr (Del Prete et al., 2017)). Despite these impressive results, further improvements are expected to be possible through ongoing developments in protein engineering and metagenomics, including protein stabilization using machine learning and de novo protein design [84], as discussed above in the context of CA for DAC.
Physicochemical properties of CA, such as solubility and viscosity, may also be amenable to optimization through protein engineering and screening of existing variants, as discussed above in the context of DAC.
Experiments could also be performed to assess how CA enhances the performance of a range of amine solvents (Hadri et al., 2016), which could be used in combination with TEAs to identify solvents that are optimal with the use of the enzyme.
3.3.3.4 CA immobilization
As discussed above for use of CA in DAC, immobilizing CA on a solid support, such as a high-wetting textile, surface-associating particle, or membrane may enable efficient catalysis of CO2 absorption while preventing the enzyme from being exposed to the high temperatures typically used in the CO2 regeneration process, which lead to enzyme degradation over time.
3.3.3.5 Solvent biofouling
As discussed for the use of CA in DAC, biofouling is also likely to be a challenge for use of CA in point-source capture. It is expected to be less of an issue for point-source capture because the main gas processing results from upstream conversion of biomass rather than air. The same approaches as discussed above can be employed for handling biofouling.
3.3.3.6 CA cost
As discussed for the use of CA in DAC, the cost of the enzyme is a challenge in point-source capture (de Oliveira Maciel, et al. 2022), but recent TEAs indicate that carbon capture using CA is cost-competitive even at current enzyme costs (see Section 3.1.3.6 (CA cost) above). Design optimizations that better leverage CA and cost reduction of CA due to increases in the scale of manufacturing are expected to further improve cost-effectiveness. The same methods for cost reduction apply here as discussed above for use of CA in DAC.
3.3.3.7 Constructing interdisciplinary teams
As discussed for the use of CA in DAC, progress on application of CA to point-source capture is limited by the limited overlap of skills and knowledge between frontier protein engineers and specialists in carbon capture. Like CA in DAC, progress on CA in point-source capture may benefit particularly from prize-based competitions if very clear targets can be defined for protein engineering of CA.
4. Recommendations
Recommendations for enabling application of CA or its mimics to DAC, DOC, and point-source capture overlap considerably because they share many constraints. While there may be substantial differences between the optimal variants of CA, deployment formats, and process designs for the different applications, most of the research methods will be similar and some projects could address multiple applications. Table 10 summarizes the relationship between the recommendations and impact on CDR, as well as the commonalities between the constraints, possible solutions, and recommendations for the three CDR methods. Recommendations for each possible solution are also summarized in Tables 3, 7, and 9 throughout this document, and are discussed in more detail below.
Table 10. Mapping recommendations to possible solutions to constraints that limit current CDR technologies.
4.1 Recommendation 1: Process modeling and TEA
Process modeling and TEA should be undertaken to characterize the potential benefits of using CA or its mimics as a (de)hydration catalyst in DAC, DOC, and point-source capture. Analyses for DAC and DOC are particularly urgent, since no such studies are publicly available, to our knowledge. This work should identify enzyme properties, enzyme deployment formats, process designs, and component designs that mutually maximize benefits, because they are interdependent. (Young et al., 2023) provides an example of similar work done to guide development of solid sorbents.
Crucially, unlike the TEAs we are aware of for CA in point-source capture (Reardon, 2015; Gilassi et al., 2020, 2021), enzyme characteristics and deployment format (such as immobilization) should be considered mutable (i.e., engineerable).
This work should provide clear targets for engineering CA and its deployment format (including pH and temperature optimum, stability, immobilization format), with clear description of stability, activity, and performance regimes that provide economic viability or diminishing returns.
In the case that CA is immobilized, it should be assumed that differently-optimized CAs and formats can be used for absorption and regeneration in DAC and point-source capture.
Care should be taken to identify whether CA or CA mimics are most appropriate for a given application. CA generally has higher catalytic efficiency, while CA mimics have greater stability.
Cost modeling should account for the substantial reduction in enzyme cost that would come naturally with economy of scale when CDR is scaled.
4.2 Recommendation 2: Estimate potential for biofouling
The degree and form of biofouling expected under various process designs should be estimated, as should its impact on cost via influence on process efficiencies and materials lifetime, as well as any added costs for de-fouling.
Care should be taken early in the process of technology-development to understand whether biofouling is likely to be a major concern for particular approaches.
Cost modeling should include the influence of biofouling whenever possible.
The impact of biofouling mitigation on process performance should be considered early in design processes.
4.3 Recommendation 3: Develop composite materials for immobilizing biocatalysts
Composite materials containing immobilized CA or its mimics should be developed for applications in DAC, DOC, and point-source capture. The needs revealed by process modeling and TEA should guide target designs as much as possible. Relevant parameters include proximity to the gas-solvent interface, catalytic efficiency, resistance to mass transport, stability of the enzyme within the material, and cost. Performance evaluations should account for expected process conditions. This work may include:
In the case of DAC and point-source capture, development of composite materials containing immobilized CA or its mimics, for example, textiles; nanoparticles or microparticles; biological CA-bearing particles such as diatom-derived particles and engineered protein particles; or membranes containing CA that:
position CA close to the air-solvent interface
are low-cost
improve the stability of CA under process conditions
optimize activity of CA
minimize enzyme leaching
enable CA immobilization at a density high enough that the rate of (de)hydration catalysis at the material surface is saturated.
In the case of DOC, development of hollow-fiber membranes containing CA.
Protein engineering to enable the above two points (for example modification of CA with fusion domains or tags that facilitate immobilization), integrated with materials engineering.
Pointed consideration of the use of immobilized CA in DAC and DOC, since it has primarily been explored for point-source capture.
Minimization of membrane resistance to mass transport for DOC, and any other method using membranes.
Platforms for cost-effective CA screening should be developed to enable faster progress in engineering CA, including platforms that simulate application conditions. This should include:
Methods for screening composite materials containing CA or its mimics in simulated application conditions, including the gas-liquid interface and flow conditions expected during application, ideally for many CAs/conditions in parallel.
High-throughput screening platforms capable of screening orders of magnitude greater than the state of the art of 103 variants/screen [38] for discovery and directed evolution of CA with target characteristics such as high catalytic rate and pH/temperature optima/stability.
4.5 Recommendation 5: Protein engineering
Protein engineering should be used to produce variants of CA, and composite materials containing CA, that meet the characteristics that modeling identifies as promising. This may include:
Improved stability and activity under process conditions.
Reduced size and/or viscosity, and increased solubility of CA under process conditions.
Localization of CA to air-solvent interface (for example, by fusion to possibly-reversible detergent peptides).
High-density and robust immobilization of CA facilitated by features of the protein, such as site-specific conjugation or binding to the material surface (e.g. via fusion of CA to cellulose- or chitin-binding domains and attachment to textiles)
Protein engineering efforts for the above work may benefit particularly from:
Recent advances in machine learning, in silico design, and de novo protein design.
Gene mining and screening of carbonic anhydrases from new or existing large-scale metagenomic sequences of polyextremophile microbial communities
In the case of immobilized CA, collaboration between protein engineers and materials scientists.
4.6 Recommendation 6: Develop low-cost biomanufacturing for CA
If the results of modeling and lab-scale efforts indicate that it is worth scaling efforts to use CA in DAC, DOC, or point-source capture, ongoing advances in protein biomanufacturing should be leveraged to produce the required CA as cheaply as possible. Such efforts are already underway for point-source capture by Novozymes.
Methods that enable flexible production of CA variants warrant particular attention. Such methods would allow experimentation with different variants at bench-top and pilot scale. Enabling scaled manufacturing of CA variants that might be required for DAC, DOC, and point-source capture will also be important. It may be helpful to incentivize biomanufacturing through market-shaping methods such as volume guarantees to enhance demand pull.
4.7 Recommendation 7: Aggressively promote community organization
Work on all the above recommendations would benefit immensely from community organization. For example:
Assembly and funding of teams or initiatives that can engage with the cross-disciplinary and interdependent nature of the above recommendations. When possible, the processes that use CA (e.g., absorption, desorption, and other processes such as reuse of particle-immobilized CA) should be developed as part of integrated efforts with protein engineering and development of composite materials containing immobilized CA, since the processes are interdependent.
Given well-defined targets for protein engineering, in silico and/or lab-based protein engineering competitions combined with centralized screening could be leveraged as a way to draw talent to the problem.
Climate tech and biotechnology should be more deeply integrated in academic bioengineering departments to enable academic work and train young researchers who will form future teams at the interface of biotechnology and climate tech.
Supplementary notes to
Note 1A: Carbonic anhydrases
The rate-limiting kinetic step in the equilibration of gaseous CO2(g) with CO2(aq) and (bi)carbonates in aqueous solution is (de)hydration of CO2 (Fig. 6A, Eqns 2 and 3). Living organisms ubiquitously confront the challenge of rapidly equilibrating CO2 and have therefore evolved enzymes—carbonic anhydrases (CAs)—to catalyze the process (Lindskog, 1997; Carbonic Anhydrases, 2019). CAs are present in all known organisms, often as multiple isozymes within the same organism (mammals have at least seven), and there are at least eight distinct families (Bose & Satyanarayana, 2017; Carbonic Anhydrases, 2019).
CAs are metalloenzymes, mostly requiring Zn2+ as a cofactor, though some use Fe2+, Cd2+, Co2+, or Mn2+[115]. The cofactor is coordinated by three histidine residues that bind and deprotonate H2O next to a hydrophobic pocket that binds CO2, and a proton is shuttled by a fourth histidine residue [116](Note 1A, Fig. 1). One cofactor-free family has also recently been discovered (Hirakawa et al., 2021). CAs are among the most efficient known enzymes: some have kcat (turnover rate) up to 4.4 x 106/s and kcat/kM up to 108 M-1s-1 near neutral pH [118], meaning they are diffusion-limited. For comparison, this means the hydration rate of CO2 in a solution of pH 14 would be achieved by ~17 µM CA, or ~0.45 g/L CA at near-neutral pH (assuming equilibrium with air, given a molecular weight of CA of 26 kDa [118], and base-promoted hydration rate of 6 x 103 M-1s-1 (Pocker & Bjorkquist, 1977; calculation here).
Given the many orders of magnitude improvements in kinetics, CA presents an exciting possibility for addressing kinetic limitations in carbon management technologies involving CO2 (de)hydration without the use of extreme conditions or toxic and costly chemicals (Bose & Satyanarayana, 2017).
An alternative catalyst to CA is a CA mimic, a non-protein molecule which mimics features of the CA active site, most importantly the amine-coordinated Zn2+ ion. An established example is the small molecule zinc cyclen (Zhang & Eldik, 1995). Frontier examples include Zn-polymer complexes and metal-organic frameworks (distinct from the MOFs with embedded CA mentioned above). (Zhang et al., 2022) describes the state of the art in CA mimics.
CA mimics are more stable than CA (Zhang et al., 2022). Because these molecules are smaller than CA, they can achieve higher concentration and lower viscosity. They may also result in less biofouling in cases when they are less accessible than proteins to microbial metabolism, which could be a significant advantage. However, they are currently either much less efficient than CA, or much more costly, or both. Improving the efficiency and cost of CA mimics is a worthy direction of research (Zhang et al., 2022). Developing methods for synthetic recapitulation of enzyme active sites may enable a breakthrough in CA mimicry.
Note 1C: Hydration kinetics at the air-solvent interface
(Stolaroff, 2013) quantifies the surface depth above which 95% of the dissolved carbon in an aqueous film is in the form of CO2(aq) (z95) for several solvents and dissolved catalysts of varying catalytic rates and concentrations (Note 1C, Fig. 1). The faster reaction kinetics of MEA compared to the carbonate solution have a clear impact on the depth to which CO2(aq) penetrates into the solvent. Notably, CA has a much faster second-order rate constant (kcat/kM) than the constants considered by (Stolaroff, 2013) and shown in Note 1C, Fig. 1, since the study was oriented toward CA mimics with lower rate constants (kcat/kM for CAs range from 105-108 M-1s-1(Mesbahuddin et al., 2021), while (Stolaroff, 2013) considered second-order rate constants from 103-105 M-1s-1).
Note 1C Figure 1: Penetration depth of CO2(aq) into a film of solvent (modified from (Stolaroff, 2013) with the author's permission). z95 is the depth above which 95% of the dissolved carbon is in the form of CO2(aq). Dotted lines represent z95 for solvents of MEA, NH3, and potassium carbonate pH 10 at undisclosed "typical" concentrations, as well as the "air-side limit", which is the z95 for a solvent that reacts with CO2 so fast that the kinetic limits of gas-side mass transfer begin to dominate. Colored lines are for solvents containing varying concentrations of dissolved hydration catalysts with the second-order rate constants shown. Note that the second-order rate constant of CA is orders of magnitude higher than the rate constants shown, since the study was oriented toward CA mimics, which are slower than CA.
Note 1D: Cost of enzyme manufacturing
By Judy Savitskaya
Fermentation notoriously exhibits strong economies of scale. That is, the cost to produce a product via biomanufacturing decreases dramatically as the scale of production increases. This effect is driven by the large fixed cost of a fermentation run through either equipment CapEx (plant build) or through contract manufacturing organization (CMO) fees plus the significant time and money required to transfer technology to a CMO. However, unlike other industries with strong economies of scale, fermentation further suffers from a variability in process performance at different scales. The optimal fermentation process that is meticulously worked out at the bench top is unlikely to be optimal as reactor sizes grow. This incoherence between lab-scale and production-scale performance, among other challenges unique to bio manufacturing (Crater, 2018), make it difficult to precisely predict the cost of scaled manufacturing.
Several ongoing efforts may help decrease protein manufacturing costs and increase the accuracy of cost projections. The most straightforward technical approach is to increase microbial production titers (grams of product per liter of fermentation broth), to effectively spread the fixed costs of fermentation across more volume of product. Novozymes and similar companies have already demonstrated protein production titers of > 100g/L in some situations [124]. There is no immediate biophysical reason that CA could not reach similar titers. A few companies claim to have genetic engineering and process development innovations that span wide classes of protein products. These claimed innovations include host modifications to support higher production of a target protein (strain engineering) and protein sequence modifications to increase expression rates (protein engineering). Another approach is to use scaled-down arrays of bioreactors to find optimal process conditions through higher-throughput experiments. Several other commercial and government-related groups are working to lower the fixed-cost portion of fermentation runs through process intensification and innovations in plant design (for example, BioMADE). Finally, while downstream processing (DSP) is often treated as an afterthought, the cost and efficiency of recovering the product from a fermentation is a major driver of ultimate cost.
Acknowledgements
We thank Matt Eisaman, Chengxiang Xiang, Sonja Salmon, Roger Aines, Noah McQueen, Jennifer Mills, Toly Rinberg, Andrew Bergman, Loren Looger, David (Doc) Brown, and Gaël Gobaille-Shaw for helpful conversations. We thank Noah McQueen, Sonja Salmon, Cara Maesano, and Frauke Kracke for helpful feedback on the manuscript. We thank Elizabeth Martindale for graphic and layout design and Stephanie Westcot for editing.
Paul Reginato is especially grateful to Sarah Sclarsic for introducing him to CDR in 2019, and to Daniel Goodwin for his outstanding friendship and partnership at Homeworld Collective.
Electrochemical pH swing methods that rely on aqueous redox carrier molecules such as water-soluble quinones can theoretically achieve lower energies (e.g., an estimated lab-scale energy of 1.1 GJ/tCO2 (Jin et al., 2020)), (Sharifian et al., 2021). However, such aqueous redox methods are at very early technology readiness levels and incur very high materials costs.
?
James Smoot:
Cite bench-scale proof of concept work https://www.sciencedirect.com/science/article/pii/S0306261917315118?casa_token=4E-Sb8C5xdEAAAAA:FLvhXQHtKWxNZX6MEVQHavASYch0uMC-kR-fthGEinUFewsYFJUO-3eeI0yymw_cCdS36QBA6IBA
Daniel Goodwin:
thanks for this! I hadn’t seen it. great citation.