Optimization and Design of the Minimal Architecture Zinc-Bromine Battery using Insight from a Levelized Cost of Storage Model

1Department of Mechanical and Aerospace Engineering, Princeton University, Princeton, New Jersey 08544, USA 2Andlinger Center for Energy and the Environment, Princeton University, Princeton, New Jersey 08544, USA 3Department of Chemical and Biological Engineering, Princeton University, Princeton, New Jersey 08544, USA 4Department of Civil and Environmental Engineering, Princeton University, Princeton, New Jersey 08544, USA

Electrical energy storage can provide many services on both the transmission and distribution side of the grid, including time-of-use energy cost management, load following, and renewables capacity firming. 1,2Of these applications, the need for renewable capacity firming is particularly pressing; 60% of the generating capacity added to the U.S. electric grid in 2016 came from intermittent, renewable sources like wind and solar. 3At present, the widespread use of electrical energy storage is limited by geological restrictions or high cost.Low-cost pumped hydro, which accounts for 94% of the storage capacity on the U.S. electric grid, suffers from considerable geological, permitting, and financing restrictions, resulting in few new installations in the U.S. within the past 25 years. 4,5][8][9][10][11][12] At present, there are two main strategies for developing electrochemical energy storage for the electric grid.The first strategy focuses on adapting existing battery technologies which were designed for other applications, like vehicles or consumer electronics.These projects usually use lead-acid or lithium-ion batteries because these technologies have decades of commercial use, resulting in relatively inexpensive capital costs.However, neither technology was designed to operate for the depth (up to 80% depth of discharge) or high number of (5,000+) cycles required to be a long-term, viable option for the grid. 2,6,9The second strategy is to implement less-mature technologies which have been particularly designed for the long-life needed on the electric grid.In particular, vanadium redox flow batteries (V-RFB) and zinc bromine redox flow batteries (ZnBr-RFB) have shown excellent performance in dozens of project plants (up to 8 MWh); however, they still maintain a high cost. 9,10,13The high cost is mainly due to the infancy of their commercial availability and the use of expensive passive components including separators, pumps, and reactors.These expensive components are used to prevent active materials from mixing, which improves performance but results in balance-of-plant and device costs far outweighing the cost of electrochemically active materials. 14herefore, a key strategy for minimizing the cost of electrochemical energy storage systems is reducing the need for expensive passive components.In this vein, Biswas et al. recently demonstrated a membrane-free, single-chamber, minimal architecture zinc-bromine battery (MA-ZBB), which has adequate performance, but with extremely low cost and relatively long cycle life. 15,161][22][23][24] Instead, separation of the bromine is obtained using a carbon foam electrode and exploiting the low miscibility of bromine to 'capture' the bromine.
In the first design of the MA-ZBB, a cell with 5 mL of electrolyte was shown to cycle with 0% capacity fade over 1000 cycles with an energy efficiency of 60% and a capacity of 40 mAh. 15This included a coulombic efficiency of ∼92%, indicating bromine self-discharge is not a "show-stopper" for the battery's development.Next, a slightly larger cell was developed (15 mL electrolyte) using the same design and including a reference electrode for performance diagnostics. 16his cell was used to identify the main reasons for the relatively low efficiency: ohmic resistances in the carbon cloth current collectors and the electrolyte.This information identified that overall cell design is crucial for advancing the performance of the MA-ZBB.
Because the MA-ZBB has no passive components, the rate of self-discharge due to bromine diffusion has the potential to be a significant problem. 25,26Therefore, operating conditions like charge/discharge time and rate are expected to greatly impact the amount of bromine self-discharge and the overall efficiency of the cell.In this work, we seek to quantify the tradeoff between rate, time, and efficiency using a scaled-up cell, which contains a total volume of 90 mL and an electrolyte volume of 70 mL.Proper quantification of this trade-off is accomplished through the development of a cost model which predicts the levelized cost of storage (LCOS) for the full MA-ZBB system (including components for power conditioning, controls, etc.).The LCOS (in $ kWh −1 cycled) is used as the figure of merit because cost is the ultimate design criteria for a grid-level energy storage system.In addition to understanding performance tradeoffs, the LCOS model is used to identify further research directions.Finally, a cost comparison between the MA-ZBB and existing technologies is conducted to demonstrate the validity of this device for grid-scale electricity storage.

Experimental
Cell construction.-Thecarbon foam electrodes (CFE) were fabricated using the same method from our previous work. 15,16The CFEs were composed of 85% carbon and 15% polyvinylidene fluoride (PVDF) by weight.The carbon was 50:50 carbon black (EC600JD, AkzoNobel) and graphite (496596-113.4G,Aldrich) by weight.The electrodes were fabricated by combining the carbon with a solution of 5% PVDF and 95% N-Methyl-2-pyrrolidone (by weight) and compressing the resulting slurry using a custom-built mold.The CFE was cured in a vacuum oven at 130 • C for 8 hours.After curing, the CFE was 4.9 ± 0.1 cm in diameter and 1.3 ± 0.1 cm tall.A detailed characterization of the electrode structure can be found in the supplemental information of. 15ll cells were constructed using a polytetrafluoroethylene (PTFE) bottle with 120 mL of total volume (<90 mL were used in all experiments).See Figure 1a for an image and schematic.The positive current collector was a thin sheet of titanium, which was cut into a circle with a small, "dagger-like" lead that was extruded out the bottom wall of the cell for electrical connection.On top of the titanium were the CFE and then a PTFE spacer made out of a 100 mL PTFE beaker (VWR, #89026-012).The spacer was constructed by removing the top and bottom of the PTFE beaker with a Dremel rotary tool.The resulting thin-walled (∼2 mm thick), hollow tube was used as the spacer.Before assembly, the spacer slightly protruded from the top of the bottle so, when the bottle lid was tightened, the spacer would compress the CFE onto the titanium for good electrical connection.The negative current collector was cut from the same carbon cloth used in our previous work. 16The cloth had a circular shape with a total area of 13 cm 2 and two, 3 cm wide strips, which were used as electrical leads by threading them through slits cut in the lid using a dremel.All open areas (i.e., slits for negative current collector leads and region where positive current collector extruded from the wall) were sealed with Gasolia pls-2 sealant containing PTFE.
Electrochemical testing.-Cycling of the cells was conducted using a MTI battery cycler.The same charge and discharge currents were used for all cycles.The energy-versus-efficiency data was collected by running the cells for five cycles at various currents.The values reported, herein, are the average value and the error bars represent the maximum and minimum deviations.
For the spacing tests, all cells were operated with a 4-hour charge and filled with 1.75 M ZnBr 2 .All experiments were conducted using the same cell, CFE, electrolyte, and current collectors.At each spacing, the cell was cycled at 80, 147, 213, and 350 mA (6.2, 11.3, 16.4, and 26.9 mA cm 2 ) with a 0.5 V cutoff.The distance between the electrodes was adjusted by stacking titanium current collectors at the bottom of the cell.The outer portion of the current collectors was sealed in Gasolia pls-2 to prevent changes in the electrochemically active surface area between tests.Energy efficiencies reported, herein, were corrected for additional resistances caused by the current collectors.
The electrolyte concentration experiments were conducted with the glass cells used in our previous work. 16In short, the cells had a total electrolyte volume of 15 mL, a negative current collector area of 4 cm 2 , and a total carbon foam mass of 1.33 g.At each concentration, the cell was cycled at 10, 20, 40, 60, and 80 mA (2.5, 5, 10, 15, and 20 mA cm 2 ) with a 0.5 V cutoff.Energy efficiencies were corrected based on deviations in the resistance of the positive and negative carbon cloth current collectors.

Results and Discussion
Electrochemical data.-Figure1a provides an image and schematic of the large cell used in this work.The cell is fabricated from a PTFE bottle and includes two major design changes from our previous, glass-cell design. 15,16The goal of the changes is to increase the energy efficiency by decreasing the ohmic resistance in the cell.The first is a "double-bussing" of the negative current collector.The second is inclusion of a thin, titanium positive current collector.Attempts to further reduce the ohmic resistance by replacing the carbon cloth negative current collector with zinc and titanium were conducted but both led to instabilities in the system (see Figures S1 and S2, respectively).More detail on the fabrication of the cell can be found in the experimental section.
The electrochemical performance of the cell is highlighted in Figures 1b-1d.Figure 1b shows the efficiency of the cell during extended cycling over six months.The data was collected with a 125 mA charge for 4 hours, followed by a 125 mA discharge to 1.0 V. Stable performance is observed for over 600 cycles with coulombic and energy efficiencies of ∼95% and ∼70%, respectively.These efficiencies are observed with no decrease in capacity, demonstrating the large cell has the same stability as the small, glass proof-of-concept. 15ote that after cycle 300, the cell experiences a decrease in energy and coulombic efficiency every ∼25 cycles.This sporadic behavior is likely caused by zinc falling off the negative current collector, which builds up because η CE is less than 100%.The reason behind the 300cycle delay is still under investigation, but we anticipate it may result from a buildup in hydrogen gas.For instance, buildup of hydrogen gas, which is generated at the negative electrode as a side-reaction, may slowly decrease the active surface area during cycling, resulting in increasingly dendritic zinc structures.The larger dendrites would be more prone to detaching from the electrode.Finally, note that the rate of zinc buildup is impacted by the rest time, where the five-minute rest between cycles does not give the cell enough time to use the selfdischarge process to reset into the discharged state.We anticipate less zinc buildup (and more consistent cycling) during normal operation where most applications will require hours or days of rest between use.At the time of publication, this test is still ongoing.
Figures 1c and 1d provide the coulombic and energy efficiencies, respectively, for cells cycled with different charge times.A separate cell is used for each charge-time with good repeatability observed between cells (Figure S3).For all charge times, as the discharge energy increases, the energy efficiency decreases.This trend is due to two factors.First, note that under our testing protocol higher discharge energies correspond to higher electrolyte utilization.Therefore, increases in discharge energy correspond to higher capacities with higher amounts of bromine in the system.The higher bromine causes more self-discharge, which results in lower coulombic efficiencies (Figure 1c) and, subsequently, lower energy efficiencies.Second, increasing discharge energy at a constant charge time is achieved by increasing the applied current.Increasing the applied current increases the resistive voltage losses, which corresponds to lower energy efficiencies.
The relative importance of these two factors depends on the charge time.For example, the energy efficiency and coulombic efficiency of the 8-hour cell decrease at average rates of 2.54% and 0.9% per Wh L −1 , respectively.These results indicate that most of the energy efficiency loss is not caused by decreases in the coulombic efficiency.Therefore, the resistive voltage losses (factor #2) are mostly responsible for the decreases in energy efficiency.On the other hand, the energy and coulombic efficiencies of the 48-hour cell decrease at rates of 0.65% and 0.5% per Wh L −1 , respectively, indicating the energy efficiency is mainly dependent on the coulombic efficiency (factor #1).
Note that the coulombic efficiencies for all cells in Figure 1c are rather high (65 to 95%) considering the results were obtained without the use of any passive components for bromine separation like membranes, complexing agents or tanks.For example, the coulombic efficiency of the 48-hour cell is >65% for all rates (up to 4000 mAh and 61% ZnBr 2 utilization).This corresponds to a self-discharge rate of ∼8.75% day −1 .The high coulombic efficiencies are attributed to the ability of the carbon foam electrode (CFE) to capture Br 2 and prevent self-discharge.For instance, the CFE demonstrates coulombic efficiencies >10% higher than commercial carbon felts in the same cell design (Figure S4).The CFE also shows reasonable capacity retention with 5-15% self-discharge after two days of standby time (Figure S5).These results suggest the MA-ZBB could be a viable option for applications requiring semi-daily, daily, and semi-weekly cycling.
Understanding the energy/efficiency trade-off at all rates is important for determining the best operating conditions for the cell.For instance, the lowest cost for energy storage is achieved at high energy efficiencies and high energy densities, which minimize the purchase and capital costs, respectively. 4According to Figure 1d, these two design criteria are inversely proportional, suggesting optimal operating conditions exist at each charge duration.For the remainder of this work, we seek to demonstrate how to exploit this energy ef-ficiency/discharge energy trade-off with a cost model and use this information to direct future research efforts on the MA-ZBB.

Model for MA-ZBB LCOS.-The figure of merit used to analyze
the performance of the MA-ZBB is levelized cost of storage (LCOS), which is the cost added to the price of electricity due to storage ($ kWh −1 ). 4 The total LCOS can be broken into three parts: i) the cost of purchasing electricity to offset storage system efficiency losses, ii) the operation and maintenance cost of the MA-ZBB, and iii) the upfront capital cost for installing the MA-ZBB (amortized over the lifetime of the battery).These three costs are accounted for in the following equation: The purchase cost (C P ) is calculated as follows: where P c is the cost of the electricity purchased on the grid ($ kWh −1 ) and η RTE is the round-trip efficiency of the whole system. 14,27In this study, η RTE is defined as: where η E is the experimentally determined energy efficiency of the device (Figure 1d), which incorporates coulombic efficiency and voltage efficiency on charge and discharge.η sys,c and η sys,d are the overall efficiencies of the system on charge and discharge, respectively, which account for losses associated with power conversion equipment. 14In this study, all charge and discharge efficiencies are identical because the cell is operated at the same rate during each step.The levelized operation and maintenance cost is determined with Equation 4: where O&M(t) is the operation and maintenance cost in $ kW −1 at each year t, η CE is the coulombic efficiency, t C is the charge time in hours, and n c (t) is the number of cycles each year.η CE is determined experimentally from Figure 1c.The summation determines the present worth of these values over the lifetime of the system, T (in years), using a discount rate, r.In this work, the O&M is constant over the system lifetime, which arises from the assumption that the MA-ZBB will be able to operate over its lifetime without any device replacement.It is also assumed the system has the same number of cycles each year.These two assumptions reduce Eq. 4 to the following: For simplicity, the number of cycles per year is calculated assuming the system is operated at the same charge rate for 90% of the year.This is captured in the following: where t Yr is the total number of hours in a year.The capital cost is broken into cost associated with i) the energy storage device and ii) power-specific components: [7]   To calculate C dev , the cost of the cell is divided by the discharge system efficiency and the discharge energy (E d ), which is determined experimentally in this work: Eq. 8 is divided by the discounted cycle life to determine the cost addition in terms of $ kWh −1 cycled.Note that E d has units of kWh; however, throughout this text, the discharge energy is plotted as discharge energy density (Wh/L) to give the reader an easily visualized number.Energy density is determined by dividing E d by the volume of the tank (V tank ).The cell components contributing to the cost are the carbon foam electrode, the zinc bromide, the current collectors, and the tank: [9]   In Eq. 9, all C's correspond to cost ($), where the superscripts ' ' and '-' refer to gravimetric (kg −1 ) and volumetric (L −1 ) quantities, respectively.The mass of ZnBr 2 is determined using Equation 10: where MW and c correspond to the molecular weight and concentration of the ZnBr 2 , respectively, while V elec is the volume of the electrolyte.
Unless specified, all cells contain 70 mL of 1.75 M zinc bromide in deionized water.Descriptions and quantities for all other values are given in Table I.The cost values in Table I were estimated from internet and literature resources. 2,4,14They represent the "best guess" for each component in the cell and are expected to fluctuate based on market and source.Therefore, these terms are reported as ranges and not absolute values.Finally, the power costs consist of balance-of-plant (BOP) and additional costs ($ kW −1 ).C power is calculated as follows: [11]   where, once again, the whole term is divided by the discounted cycle life.Similar to the work of Darling et al., the balance of plant is assumed to consist of the power conditioning equipment used to connect the DC system to the electric grid (C PCS ) and the battery control system used to control the state of charge and state of health of the device (C controls ): C add is composed of all remaining costs associated with labor, depreciation, warranty, overhead, R&D, etc. 14 All remaining constants used in the cost model of th MA-ZBB are included in Table II.The cost parameters (C PCS , C controls , and C add ) were estimated using the data provided by Darling et al.C PCS and C controls were assumed to be similar to the values reported for leadacid batteries since both systems are comprised of many aqueousbased, non-flowing cells. 14The estimated C add value for ZnBr 2 redox flow batteries was used in this work.The exact values in this work were obtained by interpolating between the estimated 2014 values and the average, projected future values reported by Darling et al.It was assumed the future values were valid for 2029, and the values reported herein were interpolated for 2018.
The O&M cost was estimated from the published O&M costs of lithium-ion and redox flow batteries ($6-$12 kW −1 ). 28,29The MA-ZBB was assumed to be 75% of that value because the MA-ZBB has no fire suppression, no HVAC requirement, and no moving parts.A discount rate of 0.07 is assumed based on the U.S. government guidelines. 30ample for determining minimum LCOS.-Figure2a provides an example cost calculation for a cell run with an 8-hour charge time, assuming an electricity purchase cost (P C ) of $0.025 kWh −1 and a lifetime (T) of 15 years. 1,4For each data point, the total, purchase, capital, and operation & maintenance costs were calculated using the discharge energy and energy efficiency data from Figure 1d as inputs.The symbols were calculated using the experimental data and the dotted lines were fits using the linear least-squares approach.First, note that as the discharge energy is increased, the capital cost decreases because the cell is delivering more energy for the same upfront cost.In comparison, as the discharge energy increases, the purchase cost increases because the efficiency of the system is decreasing (see efficiency/energy tradeoff in Figure 1d).At all energies, the O&M remains almost constant.
According to the data, there is a local minimum in the LCOS of $0.085 kWh −1 at an E d of 15.7 Wh L −1 (and η E of 57%).This minimum results from the capital cost/purchase cost trade-off described above.Figure 2b breaks down the contribution of each main cost to the total cost at LCOS min .The data indicates that C cap , C P , and C O&M account for 67%, 31%, and 2% of the total cost, respectively, suggesting that reducing the upfront capital cost while maintaining the same electrochemical performance would result in significant cost reductions.
Figure 2c further investigates this issue by breaking down the contributions of each component to the overall capital cost at LCOS min .At present, the device, BOP, and additional costs account for 60%, 17%, and 23%, respectively, indicating most of the cost comes from the electrochemical device/cell.Within cell, the CFE and electrolyte are the major contributors, each accounting for over 20% of the total capital cost.Both of these costs can be reduced through further R&D targeted at improving the electrochemical performance of the cell.For instance, at an E d of 15.7 Wh L −1 , the cell is storing ∼115 mAh/g CFE and utilizing ∼19% of the available ZnBr 2 in the solution.Improvements in both bromine capture within the CFE and utilization of the ZnBr 2 would be beneficial for reducing the cost.

Trends in LCOS and operating conditions at all charge times.-
The results in Figure 2 demonstrate that operating conditions can be selected in order to minimize the LCOS by exploiting the efficiency/energy tradeoff of the system.To understand the implication of this optimization problem across multiple charge-times and purchase costs, the calculation explained in Figure 2a was repeated for each set of data in Figure 1d, with P C values ranging from $0.02 to $0.08 kWh −1 .These values were selected to approximate the P c for a price arbitrage operation, where electricity is bought at a low value and sold at a later time at a higher value. 31The results of the calculations are shown in Figure 3 using a lifetime, T, of 15 years.Note that simulations were also conducted using lifetimes of 10 and 20 years (Figure S6).Changing the lifetime impacts the projected cost values, but does not impact any of the trends in the data or the relative importance of the major cost factors.
Figure 3a shows the LCOS min for each charge-time as a function of P C .As expected, the total cost of storage (LCOS min ) goes down as P C decreases.In addition, the projected LCOS min for the system decreases with decreasing charge/discharge time (i.e., cycling at a 4hour charge is cheaper than 48 hours).This trend occurs because of the differences in capital cost (Figure 3b), whereby the shorter charge time corresponds to lower capital costs.This behavior occurs because shorter times correspond to more cycles during the lifetime of the battery and lower self-discharge rates per cycle.This indicates that the battery cycled with shorter charge times can store more energy, more efficiently throughout its life.For instance, multiplication of E d , n cyc , and η E at P C of $0.05 kWh −1 gives values of 360, 330, 310, 230, and 140 Wh yr −1 for the 4, 8, 12, 24, and 48 hour charge times, respectively.
The fact that LCOS decreases with shorter cycles is not uncommon for an electrochemical device, especially one with appreciable self-discharge rates.What is interesting for the MAZBB is the similarity in LCOS for the 4, 8 and 12 hour cases.For instance, at a P C of $0.025 kWh −1 all three charge times have an average LCOS min between $0.079 and $0.097 kWh −1 , indicating the MA-ZBB may be a viable option for a range of applications.Applications requiring this duration time include, but are not limited to, demand charge management, time-of-use energy cost management, and renewable energy time-shift. 1 Note that the LCOS values are close to the often heralded target of $0.05 kWh −1 . 4The MA-ZBB is expected to approach this number through further optimization of the electrochemical cell.Pathways toward reaching this goal are discussed in the remaining sections of the manuscript.In addition, this number assumes the MA-ZBB can operate without failure for 5,000 to 15,000 cycles over 15 years.These numbers must be proven in the field, but are possible due to the inherent design of the system (i.e., no corrosion/passivation of active materials, no breakable mechanical parts, and no flowing liquids which increase degradation).
Figure 3b provides the contribution of C cap , C p , and C O&M to the LCOS min at each P C .For all charge/discharge times and at all purchase costs, the capital cost (C cap ) accounts for 60 to 80% of the total LCOS.Meanwhile, the purchase cost (C p ) accounts for 15 to 40% and the O&M (C O&M ), accounts for <2%.These numbers reaffirm that reducing the upfront capital cost is the most important factor for lowering the overall cost of the system.Specifically, according to Figure S7, >60% of the capital cost can be attributed to the electrochemical  I. device, were the majority of the cost comes from the CFE and the electrolyte (Figure 2c).
In Figure 3b, the relative importance of C p increases as the purchase price, P C , increases.Based on this trend, one would expect that, as P C increases, the optimal operating condition for the MA-ZBB should shift toward higher efficiencies to conserve the expensive electricity being stored.According to Figures 3c and 3d, this is precisely the case, whereby Figure 3c shows the discharge energy density and Figure 3d shows the energy efficiency of the MA-ZBB at LCOS min .As P C increases, the optimal energy decreases (Figure 3c) while the optimal energy efficiency increases (Figure 3d).
Furthermore, note what happens in the opposite direction: as P C decreases.At all rates, when P C decreases below $0.03 to $0.04 kWh −1 , the optimal energy efficiency decreases quite rapidly while the optimal energy rises with the same rapidity.This trend indicates that, no matter the operating duration (charge/discharge time), when the price of electricity is low, the most economic approach is to store as much energy as possible with little regard to the overall efficiency.The MA-ZBB is particularly suited for this cost scenario because it was designed by removing costly balance-of-plant components at the expense of a relatively low energy efficiency.

Using the LCOS model to analyze electrode spacing.-Along
with helping to understand operating conditions, the LCOS model can be used to determine the optimal design of the system which minimizes the total cost.One such design consideration is the spacing between the CFE and the negative current collector.For instance, smaller spacings between the electrodes should reduce cost since it lowers the electrolyte and tank volume in the cell.However, changing the spacing will also impact the efficiency of the cell by influencing the self-discharge and ohmic resistances.
Figure 4 shows how the LCOS model can be used to design the electrode spacing.Figure 4a shows efficiency vs. energy density for a MA-ZBB cell cycled with a 4-hour charge time with three different spacings between the electrodes: 0.4, 0.8, and 1.3 cm.The coulombic efficiency is shown in the inset.From the electrochemical data, it is unclear whether the 0.4 or 0.8 cm case will provide the lowest LCOS because neither has a clear advantage in performance at all energy densities.For instance, at energy densities >14 Wh L −1 , the 0.8 cm has the highest energy efficiency (although slightly) over all cases.For energy densities <14 Wh L −1 , the 0.4 cm case has the best energy efficiency.This occurs despite the lower coulombic efficiency with the smaller spacing (see inset), which results from more self-discharge.The better energy efficiency for the 0.4 cm case is due to the smaller ohmic resistance in the electrolyte and the smaller size of the cell.The later is influential because the data is plotted with respect to energy density.For instance, when plotting the efficiency vs. the absolute energy (as opposed to the energy density) the 0.8 cm case has the best performance due to an optimal tradeoff between self-discharge and ohmic resistance (Figure S9).
To determine the best spacing, the minimum LCOS for the three cases were calculated using the model and the results are shown in Figure 4b.For the 0.4, 0.8, and 1.3 cm cases, V tank and V elec were 66.5 and 47.0 mL; 74.9 and 55.2 mL; and 84.7 and 65.0 mL, respectively.According to the data, the 0.4 cm case provides the lowest LCOS at all P C between $0.02 and $0.08 kWh −1 .The reason for the lowest cost is demonstrated in Figures 4c and 4d which show the contributions of C cap , C P , and C O&M to the total LCOS min and the energy density at the LCOS min , respectively.According to Figure 5c, both the 0.4 and 0.8 cm cases have the same C cap , but the 0.4 cm case has better C P .This arises because the optimal energy density for both systems is <12 Wh L −1 at all P C (Figure 4d).At these energy densities, the 0.4 cm case has the superior energy efficiency.Therefore, the 0.4 cm case has the lower C P and the lower total LCOS.
Note that at all spacings in Figure 4c, C cap remains the major contributor to the total LCOS min , accounting for 60 to 80%.Amongst the capital cost contributors, the carbon foam electrode remains the major factor.It accounts for ∼26%, ∼26%, and ∼27% of the total C cap for the 1.3, 0.8, and 0.4 cm cases, respectively.In turn, the electrolyte and tank account for ∼21%, ∼18%, and ∼16% and ∼8%, ∼7%, ∼6% of the total C cap as the spacing decreases.These values indicate that decreasing tank size does not significantly impact the LCOS.In contrast, the ∼5% decrease in the electrolyte cost with decreased spacing is impactful.However, for the 0.4 cm case, the preferred operating conditions still correspond to ∼10% to ∼13% utilization of the electrolyte.Overall, these numbers suggest R&D into CFE and electrolyte optimization should still be undertaken to further reduce the MA-ZBB cost.
Determining optimal electrolyte concentration with the LCOS model.-Optimization of the electrolyte composition through changing ZnBr 2 concentration, adding supporting electrolyte, and including complexing agents is a major design consideration for maximizing the performance of the MA-ZBB.In this case study, we demonstrate how the cost model can be used to identify the best compositions by analyzing the influence of electrolyte concentration on the minimum, predicted LCOS. Figure 5 presents the results of the studies completed in glass cells. 16Glass cells were used to guarantee identical spacings between the CFE and negative current collector amongst cells.
Figure 5a provides the energy efficiency versus energy density for cells containing 0.5, 1.0, and 1.75 M ZnBr 2 .The inset shows the coulombic efficiency for the same cells.All data points represent the average of 5 cycles at a given rate, where a single cycle corresponds to a charge for 4 hours and a discharge to 0.5 V at the same rate.Note that the cell with 0.5 M has lower efficiencies at all energy densities.This is due to the lower conductivity of the electrolyte (13.2, 8.9, and 8.3 Ohm-cm for 0.5, 1.0, and 1.75 M, respectively) and the lower coulombic efficiency (see inset). 17The low coulombic efficiency is due to poor mass transport of Zn 2+ and Br − in the 0.5 M electrolyte which increases the prevalence of both hydrogen evolution at the negative electrode and oxygen evolution at the positive electrode (see Figure S8).The 1.0 M and 1.75 M cases have almost identical energy efficiencies at energy densities <10 Wh L −1 , which can be attributed to similar electrolyte conductivities.At energies >10 Wh L −1 , the 1.0 M case decreases in efficiency due to an increase in the mass transfer resistance, which corresponds to an increase in side reactions and a decrease in the coulombic efficiency.
Mass transfer resistances may present challenges in the minimal architecture design due to a lack of forced convection, which is utilized in flow batteries.At present, the optimal operating points of the cells (as determined with the LCOS model) appear to be outside the regime with considerable mass transfer resistance, which is indicated by sharp decreases in coulombic efficiency.However, these resistances must be carefully considered in all electrolyte designs.For instance, adding 1.0 M of NaCl to 1.0 M ZnBr 2 can over-support the electrolyte and cause poor cycling performance (Figure S10).
Based on the data in Figure 5a, the 1.75 M case has the best electrochemical performance at all energy densities; however, it may not be the optimal concentration because it has a higher electrolyte cost.To understand which concentration is best, the minimum LCOS for each case at purchase prices between $0.02 and $0.08 kWh −1 were calculated using the cost model.The total LCOS min is shown in Figure 5b, and the contributions from the capital, purchase, and O&M costs are shown in Figure 5c.At all purchase prices, the 0.5 M and 1.0 M cases have the highest and lowest LCOS, respectively.The high LCOS of the 0.5 M case is due to a high capital cost.The high capital cost occurs because the cell has to be operated at low energy densities due to its poor efficiency (Figure 5d provides energy density at LCOS min ).
For the 1.0 M case, the low cost is attributed to a low purchase price (C P ).For instance, both the 1.0 M and 1.75 M cases have the same capital cost at all P C , while the purchase price for the 1.0 M case is much lower (Figure 5c).This occurs because, although they have the same capital cost, the 1.0 M case reaches this cost at lower energy densities due to the lower cost of the electrolyte (Figure 5d).The lower energy densities correspond to higher η E (Figure 5a), which reduce the purchase cost (C P ).Finally, in Figure 5b, note that, as P C decreases, the LCOS of the 1.75 M case approaches the 1.0 M case.This occurs because the 1.75 M case is able to take advantage of its better η E at energy densities >10 Wh −1 .
Comparison with existing technologies.-Theprevious sections have demonstrated how a LCOS model can be used to guide optimization and design of the MA-ZBB.However, one important question not answered is whether or not the estimated, system-level costs of the MA-ZBB are competitive with existing battery technologies.To accomplish this, Figure 6 compares the upfront, capital cost and the expected round trip efficiency of the MA-ZBB with four other technologies: Li-ion (Graphite/LFP), carbon-lead acid (C-Pb Acid), vanadium redox flow (VRB), and zinc bromine redox flow (ZnBr 2 RFB) batteries.The system costs of these batteries were taken from. 14he 2018 cost was obtained by interpolating between the "2014" and "future" data in that work, assuming the "future" corresponded to 2029.The round trip efficiencies were calculated using Equation 3 and the estimated values in Table III.As mentioned previously, η E incorporates all efficiency losses occurring within the electrochemical device: round-trip coulombic efficiency and voltage efficiencies during charge and discharge.η E was calculated for the MA-ZBB as- suming the system was charged and discharged in five hours (ten hours total).The ranges of η E values for existing technologies were approximated from the literature, with emphasis placed on the most recent source. 6,9,14η sys,c and η sys,d were calculated for all technologies assuming a 4-6% loss due to the power electronics and an additional 2-4% loss for each pump.
Figure 6a provides the estimated, upfront, system-level cost of each system, including device, balance of plant, and additional costs.Note that these values ignore the amortization of the capital cost over the lifetime of the battery, which means they do not account for variations in expected cycle life of each technology.Without statistically significant commercial data, the reported lifetimes of these systems can be speculative.Therefore, following the work of Darling et al., the lifetime is not included in the technology comparison.Despite this fact, it is important to note that we expect the MA-ZBB to have a considerably longer and more reliable performance than existing technologies.For instance, both C-Pb Acid and Li-ion batteries suffer from active material passivation and corrosion, which lower their cycle life (estimations of 200-4500 and 5000-7000 cycles, respectively). 2,6,9In contrast, the aqueous-based redox flow batteries have reported lifetimes of 5,000 to 15,000 cycles.We anticipate the MAZBB to have a similar cycle life, but without future replacement costs associated with broken pumps.
Even without factoring in lifetime benefits, Figure 6a shows that the estimated MA-ZBB system cost and efficiency are competitive with other battery technologies.The solid orange rectangle is the current estimated cost based on a cell with a 0.4 cm spacing between the electrodes and a 1.0 M electrolyte.This corresponds to a 67 mL cell (with 47 mL of electrolyte) which utilizes ∼20% of the ZnBr 2 during the 5-hour charge and stores ∼70 mAh per gram of carbon in the CFE.These results were obtained with an unoptimized design of the CFE.The orange, hatched rectangle shows the projected cost of the MA-ZBB assuming further cell optimization will lead to i) a CFE which is 50-75% the current cost with the same storage capability, ii) 25% reduction in the ZnBr 2 cost through increased utilization, iii) 25-50% reduction in the current collector and tank costs due to decreased cell size, and iv) 5% improvements in round-trip efficiency.The projected costs suggest the MA-ZBB will be a low-cost, moderate-efficiency, long-life energy storage solution.Note that projected costs are not shown for other technologies because, in comparison to the MA-ZBB, these batteries are mature technologies which are not expected to have similar rapid increases in cell performance through R&D.
The results in Figure 6a include many assumptions about the balance of plant and additional costs required for each system.The values used in these assumptions are important for understanding which technology is the most cost effective.For example, the use of pumps, the HVAC requirements, and the labor needed to assemble and maintain the system all vary between these technologies, impacting the total price of the system.Along with being important, these numbers are also difficult to estimate since they depend on many variable factors, including: manufacturer, installation-site, current market, and even government legislation.
Therefore, to provide a cost comparison without these variable factors, Figure 6b shows the "cost floor" of each technology which only includes the cost of the energy conversion device (in Darling et al., this is the sum of the energy and power costs).According to the figure, the cost and efficiency of the MA-ZBB device remains competitive with projected costs of $135 to $235 kWh −1 .Note that these values are slightly higher than those of redox flow batteries.This is due to the fact that the MA-ZBB does not benefit from additional balance of plant (i.e., pumps, tanks, and pipes), which improve the energy storage capability and artificially lower the "device-only" capital cost.

Conclusions
This work presented an optimization and design analysis on the minimal architecture zinc bromine battery (MA-ZBB) using a scaled up cell which was 5-14 × larger than any previous cell reported in the literature.The scaled up MA-ZBB was shown to operate for over 600 cycles with zero capacity fade and stable energy and coulombic efficiencies of ∼70% and ∼95%, respectively.In addition, a strong tradeoff between discharge energy and energy efficiency was observed at all charge/discharge times (4, 8, 12, 24, and 48 hours).
A levelized cost of storage (LCOS) model was proposed and used to demonstrate how the energy efficiency/discharge energy tradeoff can be exploited to optimize system performance by minimizing cost.The results of the model showed that the most economical use of the battery is to operate at short cycles, which reduce the total capital cost by minimizing self-discharge.In addition, it was shown that 60% to 80% of the total LCOS came from the capital cost.A breakdown of the the capital cost showed that the electrochemical device/cell was responsible for 60% of the total capital cost contribution.Within the cell, the carbon foam electrode (CFE) and electrolyte were the most expensive components, each accounting for ∼20% of the capital cost.These results indicated that future reduction in cost of the MA-ZBB should focus on reducing the amount of carbon in the CFE and increasing the electrolyte utilization (optimum utilization in current cells is 10 to 20%), while maintaining the same electrochemical performance.
In addition, two case studies were conducted to demonstrate how the LCOS model could be used to guide the design of the MA-ZBB for low cost energy storage.One case study focused on determining the optimal spacing between the electrodes while the second study focused on identifying the best electrolyte concentration.In the first study, it was shown that a 0.4 cm spacing had the lowest LCOS at all purchase prices when compared to spacings of 0.8 and 1.3 cm.This was due to the small size of the cell, which reduced ohmic resistances and made it possible to cycle the cell at higher efficiencies while maintaining a low capital cost (the small cell was less expensive).In the second study, it was shown that a 1.0 M electrolyte provided the lowest LCOS when compared to 0.5 and 1.75 M ZnBr 2 .The 0.5 M case had the lowest upfront capital cost, but its electrochemical performance was too poor.In turn, the 1.75 M case had the highest upfront capital cost, but did not provide large enough improvement in the electrochemical performance over the 1.0 M case to justify its use in the cell.
Finally, the upfront capital costs used in the LCOS model of the MA-ZBB were compared to existing technologies.It was shown that the MA-ZBB can provide a low-cost, moderate-efficiency, long-life solution for energy storage application on the electric grid.The estimated 2018, system-level, capital cost for a fully-optimized cell operated with a 5-hour discharge was $255 to $385 kWh −1 .The device-only cost, which excluded all balance of plant and additional costs, was estimated at $135 to $230 kWh −1 .

Figure 1 .
Figure 1.Electrochemical performance of large cell.a) Image and schematic of large cell in PTFE bottle with an active volume of 90 mL and an electrolyte volume of 70 mL.b) Cycling of large cell with 1.75 M ZnBr 2 at 125 mA for a 4 hour charge and a discharge to 1.0 V. c) Coulombic and d) energy efficiency of cells operated at different charge/discharge times.Labels in d) refer to the charge time.Different energies were obtained by charging the cell at different currents.The same current was used for charge and discharge (see experimental section for details).Each data point corresponds to 5 cycles at the same current and the error bars provide the spread.

Figure 2 .
Figure 2. Cost analysis at 8-hour charge time.a) LCOS vs. discharge energy for a P C = $0.025kWh −1 and T = 15 years.Symbols were calculated from experimental data and lines were fit to data.b) Breakdown (C cap , C P , and C O&M ) of minimum total LCOS in a), which is $0.085 kWh −1 at 15.7 Wh L −1 .c) Breakdown of capital cost at minimum LCOS.

Figure 3 .
Figure 3. Minimum LCOS at all charge times.a) LCOS min at each charge time as a function of P C .b) Contribution of the main cost factors to the overall LCOS min : capital (solid, -), purchase (dash, --), operation and maintenance (dash-dot, -•).c) Discharge energy and d) energy efficiency at LCOS min .The thick lines represent the average values and the shaded regions provide the anticipated range associated with the range of cost values in TableI.

Figure 4 .
Figure 4. Determining electrode spacing using LCOS model.a) Experimental energy efficiency vs. energy density for cells with electrode spacings of 0.4, 0.8, and 1.3 cm cycled with 1.75 M ZnBr 2 at a 4 hour charge and a discharge to 0.5 V. Inset is coulombic efficiency vs. energy density for all spacings.b) Minimum LCOS for each spacing at purchase prices between $0.02 and $0.08 kWh −1 .c) Contribution of capital (solid, -), purchase (dash, --), and O&M (dash-dot, -•) LCOS to the total LCOS min .d) Energy density of each cell at LCOS min .In b-d), thick lines are average cost and shaded regions represent uncertainty associated with device cost.

Figure 5 .
Figure 5. Determining optimum concentration using LCOS model.a) Experimental energy efficiency vs. energy density for glass cells cycled with 0.5, 1.0, and 1.75 M ZnBr 2 at a 4 hour charge and a discharge to 0.5 V. Inset is coulombic efficiency vs. energy density for all concentrations.b) Minimum LCOS for each concentration at purchase prices between $0.02 and $0.08 kWh −1 .c) Contribution of capital (solid, -), purchase (dash, --), and O&M (dash-dot, -•) LCOS to the total LCOS min .d) Energy density of each cell at LCOS min .In b-d), thick lines are average cost and shaded regions represent uncertainty associated with device cost.

Figure 6 .
Figure 6.Estimated 2018 cost of grid-level batteries.Estimated upfront a) system and b) device cost for major grid-level batteries in 2018 assuming a 5-hour charge time (ignores purchase cost and amortization of capital cost).Cost values for vanadium redox flow (VRFB), lithium ion (Li-ion), carbon-lead acid (C-Pb Acid), and zinc bromine redox flow (ZnBr 2 RFB) batteries were interpolated from 14 assuming future costs corresponded to 2026.Efficiency assumptions are included in TableIII.Projected costs of the MAZBB through further cell optimization are included to provide a more realistic comparison with the other, more-mature technologies (see discussion).

Table I . Cell Cost. Values and cell parameters for terms in Equation 9.
b estimated.

Table II . Remaining Cost Values.
b η sys,c , η sys,d System efficiency on charge and discharge 0.96 b a Refs.14, 28-30.b estimated.