Abstract
Electromicrobial production technologies (EMP) aim to combine renewable electricity and microbial metabolism. We have constructed molecular to reactor scale models of EMP systems using H2-oxidation and extracellular electron transfer (EET). We predict the electrical-to-biofuel conversion efficiency could rise to ≥ 52% with in vivo CO2-fixation. H2 and EET-mediated EMP both need reactors with high surface areas. H2-diffusion at ambient pressure requires areas 20 to 2,000 times that of the solar photovoltaic (PV) supplying the system. Agitation can reduce this to less than the PV area, and the power needed becomes negligible when storing ≥ 1.1 megawatts. EET-mediated systems can be built that are ≤ 10 times the PV area and have minimal resistive energy losses if a conductive extracellular matrix (ECM) with a resistivity and height seen in natural conductive biofilms is used. The system area can be reduced to less than the PV area if the ECM conductivity and height are increased to those of conductive artificial polymers. Schemes that use electrochemical CO2-fixation could achieve electrical-to-fuel efficiencies of almost 50% with no complications of O2-sensitivity.
1 Introduction
We are moving towards a world of plentiful renewable electricity [1–3]. However, to enable high penetration of renewables onto the grid, energy storage with a capacity thousands of times greater than today’s will be essential [4–7]. Despite significant advances in electrified transportation, the need for hydrocarbons in many applications like aviation could persist and even grow for decades to come [3]. Likewise, the need to sequester tens of gigatonnes of CO2 per year will also continue to grow [8, 9]. Electromicrobial production (EMP) technologies that combine biological and electronic components have the potential to use renewable electricity to power the capture and sequestration of atmospheric CO2 and convert it into high-density, non-volatile infrastructure-compatible transportation fuels [7, 10–12].
One of the most successful demonstrations of electromicrobial production to date, the Bionic Leaf [13, 14], is capable of converting solar power to the biofuel isopropanol at efficiencies exceeding the theoretical maximum of C3 and C4 photosynthesis [15, 16]. If coupled to some of the most efficient Si or GaAs solar photovoltaics (PVs) [17], the Bionic Leaf could even outperform cyanobacterial photosynthesis, the most efficient form found in nature [18]. However, the energy storage cost of photosynthesis is ultra-low [19, 20]. Any system that aims to supplant photosynthesis will need to dramatically exceed its efficiency, its convenience and preferably both.
To date, no one has systematically explored the constraints on the efficiency of electromicrobial production systems. Here we present a model for comparing the theoretical efficiencies of systems that supply electrons to metabolism by either H2-oxidation [13, 14, 21, 22] or through a conductive extracellular matrix (ECM) by extracellular electron transfer (EET) [23]; employ in vivo enzymatic, or ex vivo electrochemical CO2 fixation [24]; and transform fixed carbon to the biofuels isopropanol [25] or butanol [20–22]. This analysis lets us calculate the maximum theoretical efficiency of each system and gives a roadmap for how to achieve it.
Theory, Results and Discussion
General Theory
Figs. 1A and 1B show simplified schematics of electromicrobial production systems with in vivo and ex vivo CO2-fixation, respectively. In Fig. 1A a microbe absorbs electricity to generate reducing equivalents needed to enzymatically fix CO2 in vivo and synthesize an energy storage molecule like polyhydroxybutyrate (PHB) or a hydrocarbon fuel. In Fig. 1B CO2 is first electrochemically reduced to a short-chain hydrocarbon like formate or formic acid ex vivo [27–29]. A microbe in the second cell absorbs electricity and further reduces and concatenates the initial fixation product to a longer-chain carbon compound. In both cases, electricity is absorbed into metabolism by either H2-oxidation (H2-mediated electromicrobial production; H2-EMP) or EET (EET-mediated electromicrobial production; EET-EMP) Fig. 1C. A complete list of symbols used in this article is included in Table S1.
We define the electrical energy conversion efficiency as the rate of energy storage molecule production, , multiplied by the energy content per molecule, Efuel, relative to the total electrical power input,
H2-mediated Electromicrobial Production is Already Optimized but can be Improved by Swapping Out CO2-fixation
Estimating the efficiency of in vivo CO2-fixation (Fig. 1A) comes down to estimating as a function of the electrical power available for electrochemistry, Pe, avail; the voltage across the electrochemical cell, ΔUcell; and the number of electrons needed to generate the NAD(P)H, Ferredoxin (Fd), and ATP for synthesis of a single fuel molecule from CO2, νef (e = elementary charge) (SI Text 1), Therefore, the overall electrical to fuel efficiency for an in vivo CO2-fixation scheme, νef can be estimated from molecular models of electron uptake. A schematic of the Ralstonia eutropha H2-oxidation machinery (used by references [13, 14, 21]) is shown in Fig. 2A. The low redox potential of H2 enables direct reduction of NADH by the cytosolic nickel-iron Soluble Hydrogenase (SH) (R. eutropha uses NADH rather than NADPH for CO2-fixation) [30, 31]. While the R. eutropha genome does not code for any Fd-reducing di-iron hydrogenases, these could be readily added to it [32–34]. Thus, the microbe simply has to oxidize a number of H2 molecules equal to the sum of NADH and Fd that it needs to synthesize a fuel molecule (the number of electrons needed is just double the number of H2).
ATP is generated by injection of electrons from H2-oxidation by the Membrane-Bound Hydrogenase (MBH) into the inner membrane electron transport chain [30, 31]; quantized energy transduction by proton pumping against the transmembrane voltage, ΔUmembrane; reduction of a terminal electron acceptor at a redox potential UAcceptor; and further quantized energy transduction by proton release through the ATP synthase and ATP regeneration.
Therefore, the number of electrons needed to synthesize a single fuel molecule through H2-oxidation is (a full derivation is included in SI Text 2) (νf, NADH, νf, Fd, and νf, ATP are the number of NAD(P)H, Fd and ATP needed for synthesis of a single fuel molecule respectively), These equations are numerically solved with the REWIREDCARBON package using estimates for the NAD(P)H, ATP and Fd requirements for isopropanol and 1-butanol synthesis (Fig. S2) from CO2 fixed by the known natural CO2-fixation cycles and the synthetic CETCH cycle [35] in Table S2.
The biggest source of uncertainty in the efficiency estimate is the transmembrane voltage (ΔUmembrane). At the time of writing we are unaware of any direct measurement of ΔUmembrane in R. eutropha or the electroactive microbe Shewanella oneidensis. Therefore, in Fig. 3 we present a range of efficiency estimates for ΔUmembrane = 80 mV (BioNumber ID (BNID) 104082 [36]) to 270 mV (BNID 107135), with a central value of 140 mV (BNIDs 109774, 103386, 109775). Counterintuitively, the efficiency of H2-mediated electromicrobial production trends downwards, moving from plateau to plateau, with increasing transmembrane voltage. (Fig. S1A). While the amount of energy stored per proton is lower at lower ΔUmembrane, energy quantization losses are also reduced.
This framework estimates the electron requirement for isopropanol and butanol synthesis by the Bionic Leaf (H2-EMP using the Calvin Cycle (CBB) for in vivo CO2-fixation) to be and respectively. The maximum electricity to isopropanol conversion efficiency of the Bionic Leaf (ΔUcell = 2 V [14]) is estimated to be (Bar C in Fig. 3). This result just exceeds the maximum reported electrical to isopropanol efficiency of 39 ± 2% [14]. This match suggests that CO2-fixation and biofuel synthesis in R. eutropha are already highly optimized.
How high could the efficiency go? Switching the product to butanol affords an improvement in H2-EMP efficiency to and a significant improvement in ease of product recovery (Bar D in Fig. 3). If the anode and cathode bias voltages could be reduced to zero, the efficiency of H2-EMP electrical to 1-butanol efficiency could rise as high as (Bar I). However, given the already low cobalt phosphate electrode overpotentials [37] in the Bionic Leaf, raising the efficiency by this route might be impractical.
Could the efficiency of EMP be increased by altering just the biological part of the system? Following intuition, electrical to fuel efficiency increases with decreasing NAD(P)H, ATP and Fd requirements for CO2 to biofuel conversion (Fig. S3A-D). The efficiencies of the six known naturally-occurring carbon fixation pathways and the synthetic CETCH pathway are shown in Fig. 3. The CETCH [35] cycle matches the efficiency of CBB (Bar E), while the naturally-occurring CO2-fixation cycles 3HP-4HB (Bar F), rTCA (Bar G) and WL (Bar H) all perform better than the Calvin cycle, raising the electrical to fuel efficiency as high as .
While the rTCA cycle and Wood-Ljungdahl pathway are both typically found in anaerobic and micro-aerophilic organisms, recent advances in compartmentalization in synthetic biology [38–40] could enable the implementation of these highly efficient pathways in synthetic organisms that operate under ambient atmospheric conditions and enable use of O2 as a metabolic terminal electron acceptor.
H2-mediated Electromicrobial Production Reaches Its Maximum Efficiency in Large Scale Systems
In principle, the efficiency of a electromicrobial production system could be independent of the specific activity of the carbon fixation pathway used (how many CO2 molecules are fixed each second by each gram of enzyme). Fixing more CO2 and storing more energy might simply require more cells operating in parallel. However, distributing electrical power through a H2 mediator could pose energetic, geometric and safety challenges [31]. To assess these challenges, we built models of H2-transport by diffusion and agitation.
The difficulty of H2-transport is determined by the number and volume of cells needed to store the H2-current, , produced by the cell current ( is the Faradaic efficiency of H2 production, typically close to 1), As hydrogenase enzymes are much faster than any carboxylating enzyme, the CO2 fixation rate is the limiting factor in electron demand per cell. The rate of electron uptake by each cell depends on the number of electrons, νef, and carbon atoms fixed, νCf, fix (not just the number incorporated, νCf), to synthesize each fuel molecule; and the rate and number of carbon-fixing enzymes, rfix and νfix (SI Text 3), Thus, the total number and volume of cells needed to store the H2-current (ncells is the cell density), H2 could be transported by diffusion from the headspace of a reactor (where it is at a partial pressure ) without any additional energy input into the system (Fig. 4A). In order to achieve the high concentration gradient needed to drive rapid diffusion of H2 ( and are the diffusion and solubility coefficients for H2 respectively), the cell culture has to be spread into a film with a height no greater than, and an area no less than (SI Text 4),
The area of a electromicrobial production system supplied by H2-diffusion scales linearly with input power while the film thickness remains the same. R. eutropha is typically grown under an atmosphere containing H2, O2 and CO2 at a ratio of 8:1:1 [42]. At the laboratory-scale, the H2 partial pressure is usually restricted to 5% of a total pressure of 1 atmosphere in order to reduce the risks of H2 explosion [42]. If supplied by a solar photovoltaic (PV), the area of the film relative to the solar PV area, APV, will remain constant. A plot of film thickness and area versus cell culture density is shown for two systems supplied by a 1 m2 solar PV in Fig. 4B: the first with a headspace H2 partial pressure of 5,066 pascals (Pa) (5% of 1 atmosphere; O2 and CO2 will both be at a partial pressure of 633.25 Pa and the system will be balanced with N2), and the second with a H2 partial pressure of 81 × 106Pa (80% of 1,000 atmospheres; O2 and CO2 will both be at a partial pressure of 10.1 × 106Pa).
For the ambient pressure system, the film area (and potential footprint of the system) is greater than the area of the PV supplying it for even the highest cell densities seen in bio-industrial applications. At the highest reported autotrophic density for R. eutropha (density region 4; n4 [43]), the film area is between 20 and 28 m2. The large film area requirement for H2-transport by diffusion at ambient pressures may not be insurmountable. Bioreactors with high internal areas but relatively small footprints could be constructed by stacking planar cell layers on top of one another, or using hollow fibers in which cells are immobilized on the walls of the fiber and reactant gases are flowed along its inner and outer surfaces [44].
Furthermore, by increasing the H2 partial pressure to 81 × 106Pa, the cell film area can be reduced to 1 m2 by a density of ≈ 5 × 1015 cells m−3, inside the range of typical cyanobacterial cell densities (density region 2; n2).
H2-diffusion systems could enable very high efficiency, but may come at the cost of high initial expenditure, complexity, maintenance, potential for H2 escape, and difficulty in removing product.
Intuitively, agitation allows H2-transport without the need for extreme system geometries, high pressures or both, at the expense of power input. The input power to the electrochemical cell is the total available electrical power, Pe, total, minus any power needed to agitate the system, We considered a cylindrical stirred tank of cells that continuously distributes H2 supplied by a sub-surface pipe (Fig. 5A). We numerically solved a set of coupled equations linking H2 production, consumption, gas transfer rate, cell culture volume, and the power required for gas mixing through an iterative algorithm in the REWIREDCARBON package using a formalism compiled by Van’t Riet [45] until a self consistent set of solutions were found (SI Text 6). The solution to these equations for a system supplied with 330W of electrical power from a 1 m2 solar PV are plotted in Figs. 5B to 5D.
At low cell densities and high system footprints (and hence volumes), the power required to transport H2 is low, while at low volumes the effort to stir is much greater (Fig. 5B). Intuitively, anyone who has grown cell culture understands that it is much easier to agitate a large cell culture (e.g. a 1 L flask) than a smaller culture (e.g. a 200 μL well in a 384-well plate). This creates a conundrum, Pe, stir can be minimized, but at the expense of a tank footprint much larger APV. Or, the tank footprint can be reduced to less than APV, but at the expense of diverting more and more solar power to mixing H2 (Fig. 5B). This means that the efficiency of the electromicrobial production system (Fig. 5C) drops precipitously from its maximum potential value to almost zero as the footprint of the system is reduced to allow it to fit under the solar PV supplying it.
The footprint-efficiency dilemma can be resolved by operating at higher input power. We calculated the system footprint to PV area ratio (Atank/APV) at which the system achieves 50%, 75%, and 95% of its maximum potential efficiency in Fig. 5D. For small scale systems (500 to 104 W of solar power) footprints of 60× to 7× the area of the solar PV supplying them are required to achieve 75% of maximum efficiency. However, for large scales systems exceeding 1.1 × 105 W of electrical power, the system footprint begins to shrink below that of the solar PV supplying it. Systems supplied by more than 1.1 × 106 W of electrical power can achieve 95% of maximum efficiency and still have a footprint smaller than the solar PV supplying them.
EET Matches the Efficiency of H2 and can Achieve High Efficiencies at Small Scales
Extracellular electron transfer (EET) could allow scale up of electromicrobial production through the use of a conductive biofilm to supply electrons to the cell (Fig. 1C). Electroactive microbes can transfer charge to, from and between external substrates like metals and even electrodes at distances up to a centimeter from the cell surface and use specialized metalloprotein complexes that connect the cell surface to the electron transport chain in the inner membrane (Fig. 2B) [46–49].
The energy landscape of EET has raised concerns about its use in electromicrobial production. The redox potentials of the membrane spanning cytochrome complex (Mtr in S. oneidensis at ≈ −0.1 V vs. the Standard Hydrogen Electrode (SHE) [50]) and the inner membrane electron carriers menaquinone (−0.0885 V [50]) and ubiquinone (0.1 V [50]) are too high to directly reduce NAD+ to NADH (−0.32 V [51]).
Nature suggests that the redox potential mismatch between the inner membrane and NAD+ is not insurmountable. Today, electroactive iron-oxidizing microbes are able to draw electrons from the oxidation of iron minerals at redox potentials from +0.7 to ≈ 0.1 V to power CO2-fixation and autotrophic metabolism [52, 53]. In the distant past it is thought that iron-oxidation powered the global carbon cycle [54]. It is speculated that an “uphill pathway” is able to lower the redox potential of electrons in the quinone pool to that of NAD+ [50].
Recently Rowe et al. [55] provided compelling evidence that a reverse electron transport chain providing an uphill pathway operates in S. oneidensis. While the the full complement of genes encoding this pathway remains unknown (although some parts have been found [55–58]), this pathway is proposed to operate by directing part of a cathodic current downhill in energy to a terminal electron acceptor and pumping protons across the inner membrane. The energy stored in the proton gradient is used to power NAD+ reduction and ATP production. A model for electron uptake by EET is shown in Fig. 2B.
Due to the need to sacrifice some current to generate a proton gradient for NAD+ (and possibly Fd) reduction, the number of electrons needed to produce the NADH, Fd and ATP for synthesis of a single fuel molecule through EET is higher than in H2-oxidation (a full derivation is included in SI Text 7), However, counterintuitively, EET-mediated electromicrobial production is not dramatically less efficient than H2-mediated electromicrobial production (Fig. 3). While the number of electrons needed to produce a molecule of fuel is higher, the whole-cell voltage in an EET-mediated system is lower than in a H2-mediated system (ΔUcell ≥ 1.23 V for H2 but only ≥ 0.92 V for EET) as the redox potential of Mtr is much lower than H2 [26]. Furthermore, the bias voltages at lab-scale remain approximately the same [7], meaning more total current is available to an EET-mediated system. However, EET-mediated electromicrobial production is approximately twice as sensitive to changes in transmembrane voltage than a H2-mediated system (Fig. S1).
The scale up of EET-mediated electromicrobial production is potentially much easier than H2-EMP. We built a model of scale up for an EET-mediated system assuming that the dominant source of overpotential is the resistivity of the biofilm. We assumed that the biofilm could be modeled as an Ohmic resistor, so that the bias voltage needed to transport electrons across it is, We developed a set of five coupled equations to solve for the cell current Icell, the bias voltage needed to drive current across the biofilm ΔUbiofilm, the area of the biofilm Abiofilm, the total number of cells in the biofilm Ncells, and the volume of the biofilm Vbiofilm in SI Text 8. These equations were solved numerically and the results shown in Fig. 6. Unlike agitation based systems, the energy cost of electron transport by EET scales linearly with system size: for a given biofilm resistivity, the ratio of the areas of the biofilm and the solar panel supplying it with electricity remains constant. Moreover, there is no obvious penalty for operating small-scale systems as there is with agitation.
At low resistivities (high conductivities) the biofilm overpotential is small, allowing a conductive matrix system to achieve close to its maximum possible efficiency, set only by the thermodynamic minimum voltages and any non-biofilm bias in the system (Fig. 6A). However, above a critical resistivity, the efficiency drops precipitously. For a 50 μm thick film, the efficiency starts to drop below 95% of maximum at a resistivity of ≈ 105 Ω cm, considerably higher than the commonly reported resistivities of Geobacter sulfurreducens and S. oneidensis biofilms (ρ4 in Fig. 6A, SI Text 9) [59–61]. Note that the peak efficiency shown in Fig. 6A exceeds that shown in Fig. 3 Bar L as we assume only anode bias.
As the resistivity of the conductive matrix increases, its thickness must decrease and its area increase in order to maintain a given efficiency. In contrast to a 50 μm film, a 1 cm thick film suffers a drop in efficiency to 50% of maximum at a resistivity of only ≈10 Ω cm, well below the resistivity range of G. sulfurreducens and S. oneidensis biofilms [59–61], but above the reported resistivities of individual S. oneidensis nanowires (ρ3 in Fig. 6A) [62] and individual filaments produced by the cable bacterium Thiofilum facile (ρ2 in Fig. 6A) [63].
Fig. 6B shows the maximum conductive matrix thickness and minimum area able to achieve a given fraction of peak efficiency as a function of resistivity. If 50% of peak efficiency is acceptable, then the biofilm area can be constrained to 1 m2 (equal to that of the solar PV supplying it) if the biofilm resistivity is 2, 650 Ω cm, well within the range of G. sulfurreducens and S. oneidensis biofilm resistivities. However, the corresponding film thickness is 440 μm, about 3× the height of most commonly observed G. sulfurreducens and S. oneidensis biofilms (although Renslow et al. did observe S. oneidensis films as thick as 450 μm). However, artificial polypyrrole conductive ECMs have been produced that are as thick as 600 μm, and have resistivities as low as 312 Ω cm (ρ5 in Fig. 6). Were the film area increased to 3.4 m2, the film thickness could be reduced to 130 μm, within the range of commonly observed G. sulfurreducens and S. oneidensis biofilm thicknesses. The biofilm resistivity would only need to be 29, 000 Ω cm, above that of many conductive biofilms, perhaps allowing some conductivity to be sacrificed to enable increased CO2 inflow or biofuel outflow.
On the other hand, if a thickness of 130 μm and resistivity of 1, 600 Ω cm are simultaneously achievable, 95% of peak efficiency can be achieved if a 6.4 m2 biofilm area is acceptable. If a 1 m2 biofilm with a resistivity 38 Ω cm and a thickness of 830 μm could be produced, 95% of peak efficiency could be achieved.
If a biofilm could be produced with a 1 cm thickness (within the range of biofilm thickness produced by cable bacteria; h3 in Fig. 6), a resistivity of 5 Ω cm (above the resistivity of individual S. oneidensis nanowires, and well above that of individual T. facile filaments, but below that of the minimum resistivity calculated by Polizzi et al. of 30 Ω cm [64]), and an area of only 0.044 m2 then 50% of maximum efficiency could be achieved. If a biofilm of 1 cm thickness, with a resistivity of 0.26 Ω cm, and an area of 0.079 m2, 95% of peak efficiency could be achieved.
Finally, if 95% of peak efficiency were desired, but only a thin biofilm of 55 μm with a high resistivity of 8, 952 Ω cm could be produced, then an area of 15 m2 would be required.
Electrochemical CO2 Fixation Could Allow Very High Electricity to Fuel Conversion Efficiencies
H2-oxidation and EET could be an important complement to electrochemical CO2-fixation technologies. Current electrochemical CO2-fixation systems typically produce compounds with no more than two carbons that are often not completely reduced [27]. By contrast, most drop-in fuels require at least 2 to 3 carbons, with 8 electrons each.
Li et al. demonstrated the reduction of formate to isobutanol and 3-methyl-1-butanol (3MB) by the H2-oxidizing microbe R. eutropha [21]. While this work relied upon oxidation of formate to CO2 and subsequent re-fixation by RuBisCO, recent advances in artificial computational metabolic pathway could enable enzymatic transformation without reliance upon this bottleneck [65, 66].
The efficiency of electrochemical CO2-fixation electromicrobial production schemes is set by the number of electrons νe, add needed to produce the NAD(P)H, Fd and ATP needed to transform the primary fixation product to a biofuel; the charge needed to synthesize the primary electro-chemical CO2-fixation product, e νep; the number of carbons in each primary fixation product, νCp; the Faradaic efficiency of the first electrochemical reaction, ξI1, (while we are calculating an upper limit on efficiency we have rarely seen ξI1 > 0.8 [27]); the efficiency of carbon transfer to the second cell ξC; and the Faradaic efficiency in the second cell ξI2 (SI Text 10), Even with only 80% Faradaic efficiency for the conversion of CO2 to formate, the electrical energy to butanol conversion efficiency of the formolase artificial metabolic pathway [65] powered by either H2-oxidation or EET exceeds all fully enzymatic CO2-fixation pathways with the exception of the rTCA cycle and Wood-Ljungdahl pathway Fig. 3, and suffers no complications of O2-sensitivity.
Conclusions
What combination of electron uptake, electron transport, and carbon fixation is the best for electromicrobial production? The model of electromicrobial production lets us sketch out a roadmap for how to proceed with the technology. We outline 10 possible development and deployment scenarios that could be pursued in the near and further future in Table 1 along with their advantages, disadvantages, and suggested niche.
This work shows that H2-EMP using the Calvin cycle [13, 14], is already highly optimized. This means that engineering the host microbe (e.g. R. eutropha) by adjusting expression levels of enzymes already encoded in the genome or changing the transmembrane voltage are unlikely to produce gains of more than a few percentage points in electricity to biofuel conversion efficiency.
One genetic engineering route to increased electrical to biofuel conversion efficiency (from ≈ 40% to as high as 55% at lab scales) is replacement of the familiar Calvin cycle with any one of the CETCH, 3HP-4HB, rTCA or WL CO2-fixation pathways. This is approach is not for the faint hearted. However, recent impressive progress in engineered the Calvin cycle into E. coli makes this a tantalizing possibility [67, 68]. Furthermore, the need to use O2 as a terminal electron acceptor to achieve maximum efficiency means that the O2-sensitivity of the rTCA and WL pathways will need to be mitigated by developing O2-tolerant versions of currently O2-sensitive enzymes in these pathways, or sequestering these enzymes inside O2-impermeable compartments inside the cell.
An alternative route to significantly enhanced efficiency is dispense with in vivo CO2-fixation and replace it with ex vivo electrochemical CO2 reduction and in vivo formate assimilation. This approach is much more genetically tractable and achieves efficiency gains comparable to replacing the Calvin cycle wih the rTCA cycle. Additionally, there is room for further improvement as new artificial pathways for processing electrochemically fixed CO2 are invented. However, this approach adds further system complexity and potential cost.
The optimization of H2-EMP with the Calvin cycle raises the question: is it time to take it out of the lab? Agitation is the most mature, lowest cost, and most easily implemented technology for electron transport considered in this article. However, the high energy cost of stirring small volumes means that the smallest increment of storage that can be built is ≈ 1 MW, about the size of a large solar farm. This is very large relative to residential storage needs (the average American home uses electrical energy at the rate of about 1.3 kW), but tiny compared to the production needs for aviation fuel (when converted to jet fuel with ≈ 50% efficiency 1 MW corresponds to ≈ 50 L hr−1. A 787-9 consumes fuel at the rate of ≈ 7, 000 Lhr−1).
Its not clear that H2-EMP will ever take on batteries for home energy storage. H2-EMP could operate very efficiently at a small power scale if H2 is transported by diffusion. However, this approach demands a high internal area reactor. This problem can be ameliorated by operating at high H2 pressure, but it is likely that this will increase cost, and incur significant safety risks. We would be foolish if we dismissed this approach outright, but we believe this analysis highlights significant technology risks.
Counter to intuition, the efficiency of EET-EMP using a reverse electron transport chain could almost match that H2-mediated electromicrobial production with laboratory overpotentials. Additionally its possible to grow conductive ECMs with sufficiently high conductivities and thicknesses that a high-efficiency, low-footprint, low internal area system could be produced with the microbes we already have available today. In principle, EET-EMP coupled to a self-assembled conductive extracellular matrix (ECM) could reduce construction costs; allow us to dispense with volatile intermediates like H2, reducing safety concerns; and allow operation in an ambient atmosphere, potentially dramatically reducing operating costs as well. Furthermore, there is no obvious penalty for operating small-scale systems, meaning that EET-EMP could enable highly distributed energy storage. However, as of today there is no easily genetically-engineered microbe capable of both electron uptake by EET and CO2 fixation, meaning that this would need to be created. It is unclear if the reductions in cost and system complexity are worth the trade-off in the amount of complex microbe engineering that would be needed for such a feat. As of today, we are unaware of the full complement of genes needed for the reverse electron transport chain. Furthermore, it is unclear how easy it would be for self-assembly of the large area ECMs that this approach would rely upon. For ECMs with conductivities similar to those produced by G. sulfurreducens and S. oneidensis several square meters of ECM would be required for every square meter of solar panel. In the lab, ECMs with areas exceeding only a few square centimeters are rarely seen [69]. If the very high reported conductivities of cable bacteria ECMs can be reproduced, these could reduce the ECM area to only a few square centimeters. Recent developments in the construction of engineered biofilms [70] suggests that it might be possible to build a biologically synthesized conductive matrix that is tailored for electrosynthesis with low resistivity, high thickness, high area, and high accessibility for CO2 and product egress.
Recent developments in coupling photo-chemistry with EET [71] opens up the possibility of constructing quantum-dot (QD)-microbe hybrids that directly inject electrons in to the EET complex and then into metabolism. This would allow for the development of a system free of photo-voltaics and electrodes that could be deployed at potentially extremely low cost. The possibility of adjusting the redox potential of the Mtr EET complex without significantly reducing efficiency (Fig. S5), along with the tunability of the electronic structure of quantum dots could allow significant room for engineering. Here, the potential for significant cost reduction could make for a significant payoff for the complex genetic engineering required to combine EET and carbon fixation.
The upper limits of efficiency of the EMP schemes presented here exceed those of all known forms of photosynthesis. Are these gains in efficiency worth pursuing? Can EMP achieve a significantly higher fraction of its theoretical efficiency in the real world than photosynthesis at an affordable cost? We cannot guarantee this, but the framework developed here gives us and other investigators the ability to rapidly understand the potential bang for buck of EMP schemes (of which there are many more than presented here). We hope that with the roadmap this framework gives, we and others in parallel can rapidly advance the field in multiple directions.
Materials and Methods
The theory presented in this work was implemented in the RewiredCarbon suite of software developed with Python with the SciPy [72] and NumPy [73] libraries. Initial visualization was implemented with Matplotlib [74]. All computer code is available at github.com/barstowlab/rewiredcarbon
Acknowledgment
This work was supported by a Career Award at the Scientific Interface from the Burroughs-Wellcome Fund (to B.B.), Princeton University startup funds (B.B.), Cornell University startup funds (B.B.), a Cornell Energy Systems Institute (CESI) Postdoctoral Fellowship (A.M.S) and by U.S. Department of Energy Biological and Environmental Research grant DE-SC0020179 (B.B.)
Footnotes
Added extra acknowledgement to Cornell Energy Systems Institute fellowship.
References
- [1].↵
- [2].
- [3].↵
- [4].↵
- [5].
- [6].
- [7].↵
- [8].↵
- [9].↵
- [10].↵
- [11].
- [12].↵
- [13].↵
- [14].↵
- [15].↵
- [16].↵
- [17].↵
- [18].↵
- [19].↵
- [20].↵
- [21].↵
- [22].↵
- [23].↵
- [24].↵
- [25].↵
- [26].↵
- [27].↵
- [28].
- [29].↵
- [30].↵
- [31].↵
- [32].↵
- [33].
- [34].↵
- [35].↵
- [36].↵
- [37].↵
- [38].↵
- [39].
- [40].↵
- [41].↵
- [42].↵
- [43].↵
- [44].↵
- [45].↵
- [46].↵
- [47].
- [48].
- [49].↵
- [50].↵
- [51].↵
- [52].↵
- [53].↵
- [54].↵
- [55].↵
- [56].
- [57].
- [58].↵
- [59].↵
- [60].
- [61].↵
- [62].↵
- [63].↵
- [64].↵
- [65].↵
- [66].↵
- [67].↵
- [68].↵
- [69].↵
- [70].↵
- [71].↵
- [72].↵
- [73].↵
- [74].↵