Aqueous Electrolytes Reinforced by Mg and Ca Ions for Highly Reversible Fe Metal Batteries

Abstract

Iron (Fe) metal batteries, such as Fe-ion batteries and all Fe flow batteries, are promising energy storage technologies for grid applications due to the extremely low cost of Fe and Fe salts. Nonetheless, the cycle life of Fe metal batteries is poor primarily due to the low Coulombic efficiency of the Fe deposition/stripping reaction. Current aqueous electrolytes based on Fe chloride or sulfate salts can only operate at a Coulombic efficiency of <91% under mild operation conditions (<5 mA/cm²), largely due to undesired hydrogen evolution reaction (HER). This work reports a series of novel Fe electrolytes, Fe electrolytes reinforced with Mg ions (FERMI) and Ca ions (FERCI), which have remarkably better Coulombic efficiency, higher conductivity, and faster deposition/stripping kinetics. By the addition of 4.5 M MgCl₂ or CaCl₂ into the baseline FeCl₂ electrolyte, the Fe deposition/stripping efficiency can be significantly improved to 99.1%, which greatly boosts the cycling performance of Fe metal batteries in both half-cells and full-cells. Mechanistic studies reveal that the remarkably improved efficiency is due to a reduced amount of “dead Fe” as well as suppressed HER. By the combination of experiments and molecular dynamics and density functional theory computation, the electrolyte structure is revealed, and the mechanism for enhanced water reduction resistance is elucidated. These novel electrolytes not only enable a highly reversible Fe metal anode for low-cost energy storage technologies but also have the potential to address the HER side reaction problem in other electrochemical technologies based on aqueous electrolytes, such as CO₂ reduction, NH₃ synthesis, etc.

Short abstract

Fe electrolytes reinforced with Mg ions (FERMI) and Ca ions (FERCI) can significantly improve the Fe deposition/stripping efficiency to 99.1%.

Introduction

Renewable energies, such as solar and wind, can decarbonize our energy generation and help to address the climate change grand challenge. However, their intermittent nature makes their integration into the grid difficult. Battery energy storage is a scalable technology that can buffer the mismatch between renewable electricity generation and grid electricity demand, but the high cost remains the main obstacle to its wide deployment. To be competitive for large-scale grid applications, battery technologies need to have a system cost of <$100/kWh and a levelized cost of storage (LCOS) of <$0.05/kWh-cycle based on the U.S. Department of Energy’s estimate.¹ To meet this goal, battery technologies need to be made of cheap and abundant materials, be easy to maintain, and have a long cycle life.

Among various battery technologies, aqueous iron (Fe) metal batteries are promising due to their low-cost potential.²⁻⁶ Fe is the second most abundant metal in the earth’s crust and is the most-produced metal commodity. Therefore, the cost of Fe ($60/ton) is much lower than any other metal used in rechargeable batteries, such as zinc ($2600/ton) and lithium ($16 500/ton).^6,7 In addition, Fe metal has a very high capacity (960 mAh/g and 7558 mAh/cm³) and outperforms zinc metal (820 mAh/g and 5854 mAh/cm³), the most popular metal anode used in aqueous batteries. Consequently, the material-level energy storage cost of an Fe metal battery is only $0.06/kWh (considering only the cost of the Fe anode), making it extremely promising for achieving the U.S. Department of Energy’s cost target for large-scale grid energy storage.¹

Motivated by this potential, many aqueous Fe metal batteries have been invented,⁸ including Fe-ion batteries,^6,9 Fe–Ni batteries,¹⁰ and all Fe redox flow batteries.¹¹ Despite these efforts, the promise of aqueous Fe metal batteries has not been realized due to their limited cycle life. Fe deposition/stripping is the designed reaction that occurs at the anode in an aqueous Fe metal battery. However, the reversibility of this reaction is far from satisfactory for making a battery with a long lifespan. In acidic aqueous electrolytes, the hydrogen evolution reaction (HER) is thermodynamically more favorable than Fe deposition (HER: −0.12 V vs SHE at pH = 2; Fe deposition: −0.44 V vs SHE). Consequently, HER competes with Fe deposition during the charging of an Fe battery. Unlike zinc metal which has a high overpotential for HER, Fe is known as a catalyst for HER;¹² therefore, HER kinetics are very facile on the deposited Fe. As a result, the Coulombic efficiency (CE) of Fe deposition/stripping is less than 91% in sulfate solutions^8,11 and less than 87% in chloride solutions¹³ under mild deposition currents (<5 mA/cm²). Such low CE leads to gas generation and electrolyte pH increase during battery cycling, which further causes the hydrolysis of Fe^2+/3+, precipitation of ferrous/ferric hydroxide, and battery performance degradation. Previous studies have shown that increasing the electrolyte pH, adding ascorbic acid buffer, chloride, or Cd additive, and raising the electrolyte temperature can increase CE.^13,14 However, the best CE reported so far is still not sufficient to build a long cycle life Fe battery. Electrolytes capable of depositing/stripping Fe metal with high efficiency (>99%) are in urgent demand for the development of long-life aqueous Fe metal batteries.

In this work, we report a series of novel aqueous electrolytes that can deposit/strip Fe metal at an efficiency of up to 99.1%, which is the best efficiency to our knowledge. The electrolytes, called Fe electrolyte reinforced with magnesium ions (FERMI) and Fe electrolyte reinforced with calcium ions (FERCI), can be simply made by adding the cheap and abundant MgCl₂ and CaCl₂ salts into the baseline FeCl₂ electrolyte. The electrochemical performance of the electrolytes is presented, and a combined experimental and computational study is performed to reveal the underlying mechanisms of the remarkable performance.

Results

Pure Fe electrolyte is made by dissolving the Fe salt into deionized water, and the FERMI is made by dissolving MgCl₂ into Fe electrolyte. We chose 0.5 M FeCl₂ (FE) as the baseline for comparing the Fe deposition/stripping efficiency and other performances of Fe electrolytes. Baseline Fe electrolytes added with different concentrations of magnesium (Mg) ions/calcium (Ca) ions are denoted as FERMI-x/FERCI-x, in which x is the molarity of the Mg/Ca ion. Fe deposition/stripping experiments were performed in Cu|Fe two-electrode cells, in which a Cu foil is the substrate for Fe deposition, and an Fe foil is the Fe source. Since the standard reduction potentials of Mg²⁺, Ca²⁺ , and Fe²⁺ are −2.37, −2.87, and −0.44 V vs SHE, respectively, Mg/Ca deposition is not likely to happen at the potential where Fe deposition occurs (>−0.5 V vs Fe/Fe²⁺). This is confirmed by the X-ray diffraction (XRD) pattern/energy dispersive spectroscopy (EDS) results (Figure 3). Typical potential curves for Fe deposition/stripping in FE and FERMI-4.5 are shown in Figure 1a, in which the negative potential corresponds to Fe deposition, and the positive potential corresponds to Fe stripping. An upper cutoff of 0.5 V vs Fe/Fe²⁺ is chosen during stripping to avoid the oxidation of the Cu substrate (Cu – 2e^– = Cu²⁺) because the oxidation potential of Cu is 0.78 V vs Fe/Fe²⁺ under standard conditions and >0.5 V vs Fe/Fe²⁺ in our studied electrolytes (Figure S1). The CE is calculated by dividing the oxidation capacity with respect to the reduction capacity. FERMI-4.5 shows not only improved CE compared to FE (99.1 ± 0.2% vs 81.8 ± 7.2%) but also increased conductivity (86.0 vs 66.7 mS/cm, Figure S2) and better deposition/stripping kinetics (total overpotential: 420 vs 900 mV). The potential curves during repeated deposition/stripping cycling are shown in Figure S3, and the CEs are shown in Figure 1b, in which an initial activation process is seen for both electrolytes. The initial cycle CE of FE is only 47.7%, whereas the initial cycle CE of FERMI is 96.8%. The FE reaches a stable CE of ∼82% after 10 cycles, whereas FERMI reaches a stable CE of 99.1% after only four cycles. This remarkably increased efficiency, 99.1%, is the best reported efficiency for Fe deposition/stripping in aqueous electrolytes to our knowledge.

Figure 3.

Characterization of the deposits in FE, FERMI-4.5, and FERCI-4.5. SEM of the deposits in (a) FE, (b) FERMI-4.5, and (c) FERCI-4.5 after the first deposition. (d) XRD of the deposits in FE, FERMI-4.5, and FERCI-4.5 after the first deposition. SEM of the Cu substrates in (e) FE, (f) FERMI-4.5, and (g) FERCI-4.5 after the first stripping. (h) EDS of the Cu substrates in FE, FERMI-4.5, and FERCI-4.5 after the first stripping.

Figure 1.

CE enhancement in FERMI (0.5 M FeCl₂ + x M MgCl₂) (a) Typical deposition/stripping voltage curves for FE (0.5 M FeCl₂) and FERMI-4.5 (0.5 M FeCl₂ + 4.5 M MgCl₂) in Cu|Fe two-electrode cells at 1 mA/cm² for 1 h. The data of the 20th cycle is shown here. (b) CE vs cycle number. (c) Average CE of FERMI-x without adjusting the pH. (d) Linear scan voltammetry of FE, 4.5 M MgCl₂, and FERMI-x electrolytes in the Cu|Fe|Fe three-electrode cell at 10 mV/s in the range of −1.0 to 0.2 V vs Fe reference electrode (RE). (e) Average CE of FE with Cl^– and SO₄^2–, the corresponding FERMI at pH = 2. Here only 3.0 M Mg²⁺ is used since the solubility of MgSO₄ is 3.2 M. (f) Average CE of A (FE), B (FE + 2.5 M FeCl₂ (3.0 M FeCl₂)), C (FE + 5.0 M NaCl), D (FE + 2.5 M MgCl₂), and E (FE + 2.5 M CaCl₂) at different pH values; 3.0 M FeCl₂ has a pH = 1, and increasing its pH leads to precipitation of Fe(OH)₂. Therefore, the efficiency of 3.0 M FeCl₂ at pH = 2 is not available. The reason to choose pH = 1 and 2 for this comparison is that Fe²⁺ will precipitate as Fe(OH)₂ when the solution pH > 2.7 (the as-made 0.5 M FeCl₂ has a pH = 2.7), and Fe³⁺ will precipitate as Fe(OH)₃ when the solution pH > 1.2 (the as-made 0.5 M FeCl₃ has a pH = 1.2).¹⁸ Note at pH = 1, there is no reversible Fe deposition/stripping in FE due to the strong acidity of the electrolyte (Figure S6), so the CE is zero.

To investigate how MgCl₂ affects the Fe deposition/stripping CE, the average CEs of FERMI at different concentrations of MgCl₂ (C_MgCl2) are compared in Figure 1c. The average CE first increases with C_MgCl2 and then starts to decrease after reaching a peak. A maximum CE, 99.1%, is achieved in FERMI-4.5. The non-monotonic dependence of CE on C_MgCl2 could be related to the non-monotonic conductivity change (Figure S2). Higher salt concentration leads to higher viscosity and lower conductivity. The associated larger overpotential leads to early termination of the stripping process, therefore lowering the CE. The low CE of FE is in large part due to the competing HER during Fe deposition, as many gas bubbles were seen on the surface of the deposited Fe during reduction. The enhanced CE in FERMI is likely due to the suppressed HER since fewer gas bubbles were seen. To confirm this, linear scanning voltammetry (LSV) tests in a Cu|Fe|Fe three-electrode cell were performed in a voltage range of 0.2 to −1.0 V in FE, FERMI, and 4.5 M MgCl₂ (Figure 1d). In FE, the peak for Fe²⁺ reduction is not visible due to the strong HER. For FERMI-2.5, the Fe²⁺ reduction peak becomes visible, and the HER current reduces. For FERMI-4.5, the HER current further decreases. The HER current in FERMI-4.5 is only −0.039A/cm² at −1.0 V, which is three times smaller than that in FE, suggesting the HER is suppressed in FERMI. For FERMI-5.3, the HER current slightly increases compared to FERMI-4.5. The dependence of HER suppression on Mg²⁺ concentration is consistent with the observed CE, confirming that the suppression of the HER is a main reason for the increased CE in FERMI. LSV results in an anodic scan showing that in the FERMI/FERCI, the Fe²⁺/Fe³⁺ redox peak has less overlap with the oxidation of the chloride solution than in the FE (Figure S4), which is beneficial for achieving high cathode CE in an all Fe flow battery.

When MgCl₂ is dissolved into FE, the pH of the solution slightly increases (pH of FE = 2.7, pH of FERMI-4.5 = 4.1) (Figure S5), which can mitigate the HER due to the reduced proton concentration. In addition, the increased concentration of Cl^– can also suppress the HER¹⁵ due to its preferential adsorption to the electrode surface.^16,17 To examine only the effect of Mg²⁺, two additional experiments were performed. First, FE made with FeCl₂ and FeSO₄ is compared with the FERMI with the same anion at the same pH (pH = 2) (Figure 1e). A clear increase in CE (28.1% for Cl^–, 15.5% for SO₄^2–) is observed when 3.0 M Mg salts are added into FE, irrespective of the anion. This result suggests that the effect of Cl^– on the CE is likely to be secondary. Second, Fe electrolytes with the same amount of Cl^– but distinct types of cations (Fe²⁺, Na⁺, and Mg²⁺) are compared at the same pH (Figure 1f). We choose a total Cl^– concentration of 6.0 M to compare the effect of different cations because the solubility of NaCl is 5.5 M and the solubility of FeCl₂ is 3.57 M. At pH = 1, the CE of FE is zero because at such a high H⁺/Fe²⁺ molar ratio (0.1 M/0.5 M = 0.2), Fe²⁺ reduction fails to compete with H⁺ reduction, so the HER dominates. The CE increases to 79.4% if an additional 2.5 M FeCl₂ is added into the FE (3.0 M FeCl₂). When comparing the electrolytes with the same Cl^– concentration (6.0 M) but different cations, electrolytes containing Mg²⁺ show better CE than those containing Na⁺ and Fe²⁺ at both pH = 1 and 2. As for the pH effect, electrolytes containing both Mg²⁺ and Na⁺ show only a slight increase (0.9–3.4%) in CE when the pH is increased from 1 to 2. In summary, these results demonstrate that (1) Mg²⁺ can enhance the CE regardless of the type of anion and electrolyte pH and (2) the significant increase of the CE in FERMI is to be primarily due to the presence of Mg²⁺ whereas pH and Cl^– only play minor roles.

A similar enhancement of Fe deposition/stripping efficiency can also be achieved with Ca²⁺. The typical potential curves for Fe deposition/stripping in FE and FERCI are compared in Figure 2a, and the CEs during cycling are compared in Figure 2b. Like FERMI, the FERCI-4.5 shows a stable CE of 98.4 ± 0.48% after the initial activation cycles. The CE of FERCI at different Ca²⁺ concentrations is shown in Figure 2c, and a maximum CE of 98.4% is achieved for FERCI-4.5. The LSV results of FE, FERCI, and 4.5 M CaCl₂ are compared in Figure 2d. FERCI and 4.5 M CaCl₂ show a similar HER suppression effect: the peak HER current at −1.0 V drops from −0.11 A/cm² in FE to −0.05A/cm² in FERCI-4.5. In addition, the dependence of HER suppression on the Ca²⁺ concentration is consistent with the CE’s dependence. These results demonstrate that Ca²⁺ can also improve Fe deposition/stripping efficiency by suppressing the HER. To confirm the enhancement of the CE in FERMI and FERCI, the CE for Fe deposition/stripping is measured with another method (Figure S7).¹⁹ As can be seen, the efficiency increases significantly after the addition of Ca²⁺ or Mg²⁺, which validates the enhancement of the CE regardless of the methods used for measuring the efficiency.

Figure 2.

Similar CE enhancement in FERCI (0.5 M FeCl₂ + xM CaCl₂). (a) Typical deposition/stripping voltage curves for FE and FERCI-4.5 in Cu|Fe two-electrode cells at 1 mA/cm² for 1 h. The data of the 20th cycle is shown here. (b) Coulombic efficiency vs cycle number. (c) Average CE of Fe electrolytes at different CaCl₂ concentrations. (d) Linear scan voltammetry of FE, 4.5 M CaCl_2, and FERCI-x electrolyte in the Cu|Fe|Fe three-electrode cell at 10 mV/s in the range of −1.0 to 0.2 V vs Fe RE.

To further understand the enhanced CE in FERMI and FERCI, scanning electron microscope (SEM) images of Fe deposits after the first deposition in FE, FERMI-4.5, and FERCI-4.5 are compared in Figure 3a–c and Figure S8. The deposits in FE are a loosely connected flowerlike assembly of nanosheets (Figure 3a, Figure S8), whereas the deposits in FERMI and FERCI are compactly stacked micrometer-sized and submicrometer-sized particles. Obviously, the deposits in FE have a larger surface area to volume ratio than those in FERMI and FERCI, which provide more sites for HER to occur. XRD of the deposits in these electrolytes is shown in Figure 3d. Strong Fe signals can be seen in all of them, suggesting the excellent crystallinity of Fe deposits. The presence of CuO and weakened Cu signals in FERMI-4.5 and FERCI-4.5 is likely a result of the chemical corrosion of the Cu substrates by Cl^–. The chemical corrosion of Cu in chlorine solution is well-known²⁰ and can be described by Cu + H⁺ + 2Cl^– = 0.5H₂ + CuCl₂^–. Note the electrochemical oxidation of Cu cannot explain this phenomenon because the working electrode is in a reductive environment during the Fe depositing process. The chemical corrosion does not intervene in the calculation of the Coulombic efficiency because the electron transferred from Cu to H⁺ does not go through the external circuit; therefore, it is not counted by the potentiostat during the Fe deposition/stripping experiment. SEM images of the Cu substrates after the first stripping are compared in Figure 3e–g and Figure S9. Many fluffy clusters of a few micrometers in size and large quantities of nanoparticles exist on the surface of the Cu substrate in FE. Its EDS shows a strong signal of Fe Lα (Figure 3h, Table S2), suggesting these clusters and nanoparticles are unreacted Fe during the stripping process, i.e., “dead Fe”. Such “dead metal” is commonly observed in the stripping of electrochemically deposited metals, such as Li and Na.²¹ It occurs when metal deposits are electrically isolated from the substrate during the stripping process, which is common for deposits with skinny morphology.²² The weak signal of Cu Lα indicates that the dead Fe covers most of the Cu substrate. The strong O Kα1 signal is likely due to the oxidation of the dead Fe during the sample preparation. In stark contrast, far less dead Fe is observed on the Cu substrates in FERMI-4.5 and FERCI-4.5, and EDS shows a strong sign of Cu Lα but no clear sign of Fe Lα, confirming there is little dead Fe on the Cu substrates. The strong signal of Cu and the absence of anion signals (Cl and O) also suggest that no solid electrolyte interface (SEI) forms on the Cu substrate since SEI should contain compounds of the corresponding ions in the electrolyte. Mg and Ca or their oxides/hydroxides are not observed in either the XRD or EDS results, confirming that Mg/Ca deposition does not occur in these electrolytes. In summary, these results demonstrate (1) only Fe deposition occurs in FERMI and FERCI, (2) they promote the growth of large and compact Fe deposits, and (3) they reduce the amount of dead Fe during the stripping process. The better Fe deposits morphology and less dead Fe in FERMI and FERCI are other reasons for the better Fe deposition/stripping efficiency.

To demonstrate how the Fe deposition/stripping efficiency affects the Fe metal battery’s cycle life, cycling experiments were performed with Fe|Fe symmetrical cells and LiFePO₄|Fe full-cells (Figure 4). During cycling, side reactions, including HER, will change the electrolyte’s chemistry over time and eventually fail the cell. The Fe|Fe symmetrical cells were cycled by charging for 0.1 h and then discharging for another 0.1 h both at 1.0 mA/cm² without constraining the voltage. The voltage profiles during cycling are compared in Figure 4a and b. The cells with FE can work for 88.5 h, and then the cells fail, which is signaled by a sudden increase in voltage caused by an increase in the internal cell resistance from 0.86 to 17.1 Ω (Figure S10). The failed coin cells swelled, indicating the generation of a large amount of gas inside the cell (Figure S11). After the failed cells were assembled, the electrolyte almost dried out, and green precipitate could be found on the spacer (Figure S12). These results suggest the internal resistance increase can be attributed to the following reasons: (1) the generated gas bubbles block the ion transport pathway between the working and counter electrodes, (2) HER consumes water and leads to an increase in electrolyte viscosity, and (3) the pH increase and the precipitation of Fe salts. In contrast, the cell with FERMI-4.5 and FERCI-4.5 can work for >250 h with no significant increase in voltage. Similar results were observed in Cu|Fe cells as well (Figure S13). To further demonstrate how the improved Fe anode CE affects the Fe metal battery’s cycle life, LiFePO₄|Fe full-cells with FE, FERMI-4.5, and FERCI-4.5 are assembled and tested. The cells were charged/discharged at 1.0 mA/cm² in the voltage range of 0.60–1.25 V. The voltage profiles during cycling are compared in Figure 4c. The cell with FE fails rapidly within the first 20 h, whereas the cells with FERMI-4.5 and FERCI-4.5 can operate for over 80 h without clear signs of degradation. The normalized capacity and CE are compared in Figure 4d and e, respectively. All cells undergo an activation process before reaching the maximum capacity, which could be attributed to the slow wetting of aqueous electrolytes to the graphite felt current collector. The cell with FE fades rapidly and loses 97.7% of the capacity at the 100th cycle, whereas cells with FERMI-4.5 and FERCI-4.5 show very stable cycling and lose only 11.2% and 4.67% of the capacity at the 100th cycle. Meanwhile, the cell with FE shows an average CE of 95.2%, whereas the cells with FERMI and FERCI show average CEs of 97.7% and 97.2%, respectively. The cycling performances and CEs of LiFePO₄|Fe full-cells in FE, FERMI-4.5, and FERCI-4.5 with no Li salts are given in Figure S14. Cells with FERMI-4.5 and FERCI-4.5 show much more stable cycling than cells with FE, which is consistent with Figure 4d and e. In summary, these results demonstrate that the enhanced Fe anode efficiency in FERMI and FERCI can significantly boost the cycling performance of Fe metal batteries in both half-cells and full-cells.

Figure 4.

Cycling performance of Fe|Fe symmetric cells and LiFePO₄|Fe full-cells in FE, FERMI-4.5, and FERCI-4.5. (a) Voltage vs cycling time of Fe|Fe symmetric cells with FE, FERMI-4.5, and FERCI-4.5. Cycling condition: 1.0 mA/cm²_. (b) The first and last 4 h of the cycling results. (c) Voltage vs cycling time of LiFePO₄|Fe cells with FE, FERMI-4.5, and FERCI-4.5. Cycling condition: 1.0 mA/cm². (d) Normalized capacity and (e) CE vs cycle number of LiFePO₄|Fe cells with FE, FERMI-4.5, and FERCI-4.5.

To understand the effect of Mg²⁺ and Ca²⁺ on the Fe electrolytes, Raman and Fourier transform infrared (FTIR) spectra of FE, FERMI, and FERCI are collected and compared (Figure 5). In aqueous electrolytes, the water interacts strongly with the ions by electrostatic interaction, H-bonding, or charge transfer.²³ The preferred orientation, H-bonding, and vibrational dynamics of water in the hydration shell are very different from those of the bulk water. Water can form a maximum of four H-bonds with its neighboring water molecules by donating two protons and accepting two protons to the lone pair of electrons on oxygen. Based on how strongly a water molecule participates in H-bond formation, four types of water are possible at ambient temperature. In the order of decreasing number of H-bonds, they are DDAA, DDA, DAA, and DA, in which D refers to water molecules donating a proton, and A refers to water molecules accepting a proton (i.e., DDA means double donor–single acceptor).²⁴ In the Raman spectra of pure water, the OH symmetric stretch region has a broad peak with three bands at ∼3200, ∼3400, and ∼3600 cm^–1, which can be assigned to DDAA water, DA water, and DDA water, respectively.²⁴ In FE, the molar ratio of H₂O to Fe²⁺ is 105 (Table S1). Since Fe²⁺ prefers octahedral coordination, there are at most six water molecules in its hydration shell so that most water molecules exist in the bulk. Therefore, the Raman spectroscopy of FE is close to that of pure water. Nonetheless, the water structure changes significantly after the addition of Mg²⁺ or Ca²⁺. Due to the high charge density of Mg²⁺ and Ca²⁺, its influence on water structure and dynamics extends beyond the first hydration shell. In dilute electrolytes, both Mg²⁺ and Ca²⁺ immobilize ∼20 water molecules,²⁵ forming two hydration shells around them, with the first hydration shell containing six water molecules for Mg²⁺ and six–nine water molecules for Ca²⁺.²⁶ Upon an increase of the concentration of Mg²⁺/Ca²⁺, the number of water molecules in the hydration shell of Mg²⁺/Ca²⁺ increases proportionally up to a certain concentration (2.0 M for Mg²⁺, in which the molar ratio of water and Mg²⁺ is ∼25).²⁷ With further increase of the concentration of Mg²⁺/Ca²⁺, solvent-separated ion pairs (2SIP) or even solvent-shared ion pairs (SIP) can form. At 4.5 M Mg²⁺/Ca²⁺, the molar ratio of H₂O to Mg²⁺/Ca²⁺ is ∼10, suggesting that all water molecules exist in the hydration shells of Mg²⁺/Ca²⁺, and a portion of this hydration shell water is shared with Cl^–. In the hydration shells, the O atoms of water point toward Mg²⁺/Ca²⁺, and H atoms point away. For water molecules shared with Cl^–, their H atoms point toward the Cl^–. Due to this orientation preference and geometric constraint, hydration shell water forms fewer H-bonds than bulk water. Therefore, the presence of a large amount of Mg²⁺/Ca²⁺ disrupts the water structure and eliminates strongly hydrogen-bonded water. Previous studies show that the ∼3600 cm^–1 band and ∼3200 cm^–1 band in the Raman spectrum weaken as 1.0–2.0 M Mg²⁺/Ca²⁺ is added.^27,28 Here in our results, a similar weakening effect is observed at 2.5 M of Mg²⁺/Ca²⁺. These two bands completely disappear at 4.5 M Mg²⁺/Ca²⁺, and further increase of the Mg²⁺/Ca²⁺ concentration results in no observable change (Figure 5). These results indicate that the addition of 4.5 M Mg²⁺/Ca²⁺ eliminates DDA water and DDAA water, which leads to fewer H-bonds per water. Similar suppression of the ∼3200 cm^–1 band is also observed in the OH-stretching region of the FT-IR spectra of FERMI and FERCI. The intensity of the O–H–O bending vibration peak at 1600 cm^–1 grows with increasing Mg²⁺/Ca²⁺ concentration, also indicating the weakening of the H-bond. In addition to the change in water structure, the hydration shell of Fe²⁺ also changes after adding Mg²⁺/Ca²⁺. In FE, Fe²⁺ mostly exists as [Fe(H₂O)₆]²⁺ in an octahedral configuration.²⁹ Given that the ratio of Fe²⁺/H₂O decreases in FERMI and FERCI, the solvation shell of Fe²⁺ will have fewer water molecules and more Cl^–. A similar effect has been observed for Zn²⁺ when less water is available.³⁰ The computational study below will elucidate this change in the Fe²⁺ solvation shell.

Figure 5.

Raman spectra and FT-IR Raman spectra of (a) water, FE, FERMI-x, and 4.5 M MgCl₂ and (b) water, FE, FERCI-x, and 4.5 M CaCl₂. FT-IR of (c) water, FE, FERMI-x, and 4.5 M MgCl₂ and (d) water, FE, FERCI-x, and 4.5 M CaCl₂.

To further understand the electrolyte structure, atomistic molecular dynamics (MD) simulations using the polarizable force field (APPLE&P)^31,32 were performed for FE (0.5 M FeCl₂) and FERMI-4.5 (0.5 M FeCl₂ + 4.5 M MgCl₂) at room temperature. The simulations contained 4000 water molecules and the corresponding number of ions. The snapshots in Figure 6a illustrate the structure of both electrolytes. In 0.5 M FeCl₂, most water molecules do not interact with ions. While the ion can form small clusters, they are homogeneously distributed throughout the system. In 0.5 M FeCl₂ + 4.5 M MgCl₂, the electrolyte structure and distribution of water change significantly. Figure 6a shows that ions now form a continuous phase, and the water structure is significantly perturbed. The cation–oxygen of water (O_w) radial distribution functions (RDFs) with corresponding apparent coordination numbers are shown in Figure 6b as a function of shell radius, and the cation–anion RDFs are shown in Figure 6c. The strong first peak in the RDF defines the first solvation shell (3.40 Å for Fe²⁺ and 3.15 Å for Mg²⁺). The shorter Mg–O_w distances and stronger first and second solvation shell peaks of Mg²⁺ indicate that water molecules are tightly bound to Mg²⁺ due to stronger charge localization. The shorter Mg–Cl distance than Fe–Cl distance reveals a closer packing between Mg²⁺ and Cl^– than for the Fe²⁺. The numbers of water molecules and Cl^– in the solvation shell of Fe²⁺ and Mg²⁺ are given in Table S3. The first solvation shell of Fe²⁺ contains an average of 4.5 water molecules and 1.6 Cl^– in 0.5 M FeCl₂. After the addition of 4.5 M MgCl₂, the solvation shell of Fe²⁺ changes to 3.0 water molecules and 3.1 Cl^–. The first solvation shell of Mg²⁺ contains an average of 3.4 water molecules and 2.4 Cl^–.

Figure 6.

Electrolyte structure from MD simulations: (a) snapshots of 0.5 M FeCl₂ (left) and 0.5 M FeCl₂ + 4.5 M MgCl₂ (right) from MD simulations. Ions are highlighted as spheres. O_w atoms in the first coordination shell of the cation are highlighted as red spheres. Snapshot on the far-right highlights the water distribution in which all ions are shown as the semitransparent blue background. (b) The cation–O_w radial distribution function (left y-axis) and the corresponding coordination number (right y-axis). (c) The cation–anion radial distribution function (left y-axis) and the corresponding coordination number (right y-axis). (d) Hydrogen bonding. The water–water H-bond is determined by three criteria, i.e., r^OH < 2.45 Å, θ^HOO < 40°, and two water molecules located inside the first coordination shell. The colored pixels illustrate the probability of water molecules participating in different numbers and types of H-bonds. N_A is the number of H-bonds water participates in by accepting a proton, and N_D is the number of H-bonds water participants in by donating a proton.

Analysis of water–water hydrogen bonding (Figure 6d and Table 1) shows that in 0.5 M FeCl₂, water molecules on average have 3.2 H-bonds, with 36.6% of them participating in four or more H-bonds (donor and acceptor combined), 32.1% participating in three H-bonds, 23.1% participating in two H-bonds, and only a small fraction participating in one or no H-bond. The addition of 4.5 M MgCl₂ significantly perturbs the H-bonding network between water molecules, as the average number of H-bonds per water molecule reduces to 2.2 and the fraction of water molecules that participate in four or more hydrogen bonds drops to 20.0%. Instead, the fraction of water molecules participating only in one or no H-bond increases to more than 32%. Note, in molecular simulations, depending on the geometric definition of the H-bond, molecules that have more than four H-bonds are possible. This is because no matter what definition one chooses, there will always be some molecules in transition between two H-bonds where both bonds will formally fall within the boundary of the definition and be counted.³³ For the same reason, the computed average number of H-bonds per water molecule will be higher than in experiments. In addition to the reduced number of H-bonds per water, the averaged H-bond length increases from 2.024 to 2.085 Å after the addition of MgCl₂, indicating the weakening of the H-bond strength.

Table 1. Number of H-bonds per Water Molecule and the Corresponding Probability.

In summary, the simulations show that the electrolyte structure undergoes several major changes after the addition of 4.5 M MgCl₂: (1) Mg²⁺ is strongly bound with water molecules; (2) the number of H-bonds per water molecule reduces from 3.2 to 2.2; (3) the number of water molecules in the first solvation shell of Fe²⁺ reduces from 4.5 to 3.0; and (4) the average length of H-bonds increases. These changes are consistent with our analysis based on the Raman spectroscopy and FT-IR results. With fewer H-bonds per water molecule and longer H-bonds, the water O–H covalent bond becomes shorter and stiffer, therefore making hydrogen evolution more difficult.^34,35 In addition, the reduced number of water molecules in the hydration shell of Fe²⁺ makes water reduction more difficult because the likelihood of water reduction decreases when Fe²⁺ is brought to the vicinity of the electrode surface for the deposition reaction.^36,37

To further elucidate the water molecule’s enhanced resistance to the reduction in FERMI-4.5, density functional theory (DFT) calculations were performed to calculate the reduction potentials of water in the solvation shell of different cations and water with different numbers of H-bonds. First, we investigated the reduction of MCl₂ clusters (M = Fe²⁺, Mg²⁺, and Ca²⁺) hydrated with five water molecules and one hydronium ion. The latter is introduced due to the acidic environment of the investigated electrolytes. Considering the first-electron reduction reaction generating the hydrogen radical as the rate-limiting step, the calculation of absolute reduction potential has been achieved by using a traditional Born–Harber cycle (Figure 7a and b), which is widely used in the calculations of redox reactions for battery electrolytes and electrochemical reactions in aqueous phases.³⁸⁻⁴⁰ When Fe²⁺ is the cation, the reduction potential is −0.302 V vs Fe/Fe²⁺, but the reduction potential reduces to −0.999 and −0.723 V when the central cation is replaced with Mg²⁺ or Ca²⁺, respectively. Next, we investigated the influence of the number of H-bonds per water molecule on its reduction. Our experiment and MD simulation both show that the average number of H-bonds per water decreases after the addition of 4.5 M MgCl₂. To examine how the number of explicit H-bonds that water molecules participate in affects its reduction potential, we compared the reduction of water molecules with four and two H-bonds. A water molecule with four H-bonds (2A2D) is represented by the central H₂O in a five-H₂O cluster (Figure 7c). Water with two H-bonds is represented by the central H₂O in a three-H₂O cluster. Since there are three different isomers of three-H₂O clusters with 1A1D, 0A2D, and 2A0D H-bonds, the reduction potentials of the central water in them were computed respectively to investigate how the H-bond type influences the reduction potential. Using the reduction potential of the central H₂O in a five-H₂O cluster as the reference, the reduction potentials of the central H₂O in the three-H₂O clusters were found to be −0.565, −0.654, and −0.544 V for 2A0D, 1A1D, and 0A2D, respectively. The conducted DFT calculations demonstrate that (1) the presence of Mg²⁺ and Ca²⁺ makes water more reduction resistant, and (2) decreasing the number of H-bonds that water participates in also makes water molecules more reduction resistant.

Figure 7.

Electrochemical reduction of H₂O. (a) The calculation of the electrochemical reduction potentials of MCl₂ clusters (M = Fe²⁺, Mg²⁺, and Ca²⁺) hydrated with five water molecules and one hydronium ion. (b) Geometries of the hydrated FeCl₂ cluster investigated for the reduction reaction. (c) Geometries and reduction potentials of water molecules in a cluster with different numbers and types of H-bonds. The reduction potential of three-H₂O clusters is given relative to the five-H₂O clusters.

Based on the above experimental results, the enhanced CE in FERMI and FERCI can be attributed to two reasons (Figure 8). The first is less dead Fe during the stripping process. The presence of Mg²⁺/Ca²⁺ leads to Fe deposition with larger particle size and smaller surface area, which tends to form less dead Fe during stripping. The second is the suppression of HER because water molecules become more reduction resistant. The combined experimental and computational study suggests that the enhanced water stability toward reduction is because (1) water molecules are tightly bound by Mg²⁺/Ca²⁺ in their hydration shells; (2) Mg²⁺/Ca²⁺ significantly disrupts the H-bond network of water by reducing the number of H-bonds per water and increasing H-bond length, therefore strengthening the covalent O–H bond of the water molecules;⁴¹ and (3) the reduced water concentration results in fewer water molecules in the hydration shell of Fe²⁺, which lowers the chance of water reduction when Fe²⁺ is brought to the electrode for Fe deposition.⁴²

Figure 8.

Schematic of (a) the structure of pure Fe electrolyte and (b) the structure of FERMI and FERCI. The improved CE is due to the suppression of HER and less dead Fe.

In addition to the remarkable enhancement of CE for Fe deposition/stripping, this study also reveals several interesting discoveries whose understanding could benefit the development of aqueous electrolytes for electrochemical technologies. An increase of the concentration of FeCl₂ alone from 0.5 to 3.0 M can increase the CE from 0% to 79.4% (at pH = 1). Theoretically, the addition of more Fe salts to the Fe electrolyte in the water-in-salt regime could further stabilize water and enhance the CE, as demonstrated in Zn electrolytes³⁶ and Li electrolytes.³⁷ However, the limited solubility of common Fe salts (FeCl₂: 3.57 M, FeSO₄: 3.0 M) makes it impossible to explore this regime of Fe electrolytes. This challenge may be tackled with highly water-soluble salts based on organic anions such as bis(trifluoromethanesulfonyl)imide (TFSI^–). The presence of a high concentration of Cl^– can enhance Fe deposition/stripping CE in electrolytes with NH₄⁺,¹⁴ but this appears only as a surface effect. This is because Cl^– can preferentially adsorb on the Fe electrode^17,37 but imposes a relatively smaller perturbation on the water structure.^23,43 Such surface effects of Cl^– are dwarfed by the bulk electrolyte structure change when cations like Na⁺, Mg²⁺, and Ca²⁺ are added due to their strong ability to bound water.²³ Because Na⁺ can also reduce the number of water molecules in the hydration shell of Fe²⁺ and disrupt the water structure,⁴³ the addition of Na⁺ to FE also enhances the CE (Figure 1f), albeit to a lesser extent compared to Mg²⁺ and Ca²⁺. Such supremacy of Mg²⁺ and Ca²⁺ over the monovalent Na⁺ is linked to their stronger ability to bind water and their ability to perturb unoccupied molecular orbitals of hydration water.²³ Between Mg²⁺ and Ca²⁺, Ca²⁺ is slightly less effective than Mg²⁺ in enhancing the CE, which can be explained by its slightly weaker hydration than Mg²⁺ and its lesser extent of perturbing the hydration water orbital.^23,25 The effect of Mg²⁺ and Ca²⁺ in suppressing HER seems universal, as our preliminary results show that they can also enhance Zn deposition/stripping efficiency, but this discussion is beyond the scope of this work and will be presented in a future publication. Lastly, when the proposed electrolytes are used in an Fe metal battery with an intercalation cathode, such as the Prussian blue analogue,⁶ Mg²⁺ or Ca²⁺ can also insert into the cathode. The selectivity of the intercalation reaction toward the alkaline earth metal ion and Fe²⁺ requires further investigation, which is beyond the scope of this work. However, this is not a concern for Fe metal flow batteries, in which charge/discharge involves electron transfer from/to the soluble redox-active ions or molecules.

To further increase the efficiency, many possible strategies can be adopted.⁴⁴ One can introduce some surface film-forming components (additives, cosalts, or cosolvents) into the aqueous electrolyte. This method is widely used in aqueous Li-ion batteries^37,45 and Zn-ion batteries.^46,47 The irreversible decomposition of these components may form an Fe²⁺ conductive but electron-insulating film (termed solid-electrolyte interphase, SEI), which allows the deposition of Fe but prevents further decomposition of water, therefore improving the efficiency. Another strategy is to suppress HER by reducing the water concentration, which can be done by further increasing the salt concentration or adding an organic solvent.⁴⁸ HER can also be suppressed by increasing the HER overpotential, which can be achieved by introducing anticatalysts such as Bi onto the Fe metal electrode. In addition, the morphology of the Fe deposits can be further improved by adding surface surfactants to reduce the amount of dead Fe.

Safety Statement

No unexpected or unusually high safety hazards were encountered.

Conclusion

In conclusion, electrolytes that can support highly reversible Fe metal anode are critical to realizing the potential of aqueous Fe metal batteries as a low-cost energy storage technology. Two novel aqueous electrolytes, FERMI and FERCI, are reported in this work, and they show remarkably better Fe deposition/stripping efficiency (99.1%), higher conductivity, and lower overpotential than the baseline Fe electrolytes. Both half-cell and full-cell studies show that batteries with the baseline Fe electrolyte fail very quickly because the HER leads to large internal resistance, whereas batteries with FERMI and FERCI show significantly better cycling stability, which demonstrates the potential of these electrolytes for realizing long-cycle Fe metal batteries. Comprehensive experimental and computational studies reveal that the enhanced Fe deposition/stripping efficiency is due to a synergy of improved deposit morphology (therefore less dead Fe) and enhanced water reduction resistance. Due to the simple fabrication method and low cost of raw materials, these novel electrolytes are ideal for unleashing the low-cost benefit of Fe metal batteries, especially Fe flow batteries. Broadly, the novel electrolytes reported here not only enable long-cycle Fe metal batteries but also open a new avenue to address the HER side reaction for other electrochemical technologies based on aqueous electrolytes, such as the CO₂ reduction, NH₃ synthesis, etc.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscentsci.2c00293.

• Experimental details, SEM images, cycling performance, and electrochemical property measurements (PDF)

Author Present Address

^§ Chemical Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831-6119 United States

The authors declare no competing financial interest.