Lithium–antimony–lead liquid metal battery for grid-level energy storage

All-liquid batteries comprising a lithium negative electrode and an antimony–lead positive electrode have a higher current density and a longer cycle life than conventional batteries, can be more easily used to make large-scale storage systems, and so potentially present a low-cost means of grid-level energy storage. The integration of batteries into the electric grid is seen as possible means of regulating energy supply from intermittent sources such as wind or solar, but today's battery technologies are too expensive to do the job. An all-liquid battery, comprising a liquid negative electrode, a molten salt electrolyte, and a liquid positive electrode, is one of the technologies being investigated for this role. Here Kangli Wang and colleagues describe a new variant of the concept — an all-liquid Li||Sb–Pb battery — that, through careful choice and alloying of the component electrode materials, reduces operating temperatures and hence potential cost while retaining desirable performance characteristics. The ability to store energy on the electric grid would greatly improve its efficiency and reliability while enabling the integration of intermittent renewable energy technologies (such as wind and solar) into baseload supply1,2,3,4. Batteries have long been considered strong candidate solutions owing to their small spatial footprint, mechanical simplicity and flexibility in siting. However, the barrier to widespread adoption of batteries is their high cost. Here we describe a lithium–antimony–lead liquid metal battery that potentially meets the performance specifications for stationary energy storage applications. This Li||Sb–Pb battery comprises a liquid lithium negative electrode, a molten salt electrolyte, and a liquid antimony–lead alloy positive electrode, which self-segregate by density into three distinct layers owing to the immiscibility of the contiguous salt and metal phases. The all-liquid construction confers the advantages of higher current density, longer cycle life and simpler manufacturing of large-scale storage systems (because no membranes or separators are involved) relative to those of conventional batteries5,6. At charge–discharge current densities of 275 milliamperes per square centimetre, the cells cycled at 450 degrees Celsius with 98 per cent Coulombic efficiency and 73 per cent round-trip energy efficiency. To provide evidence of their high power capability, the cells were discharged and charged at current densities as high as 1,000 milliamperes per square centimetre. Measured capacity loss after operation for 1,800 hours (more than 450 charge–discharge cycles at 100 per cent depth of discharge) projects retention of over 85 per cent of initial capacity after ten years of daily cycling. Our results demonstrate that alloying a high-melting-point, high-voltage metal (antimony) with a low-melting-point, low-cost metal (lead) advantageously decreases the operating temperature while maintaining a high cell voltage. Apart from the fact that this finding puts us on a desirable cost trajectory, this approach may well be more broadly applicable to other battery chemistries.

The ability to store energy on the electric grid would greatly improve its efficiency and reliability while enabling the integration of intermittent renewable energy technologies (such as wind and solar) into baseload supply [1][2][3][4] . Batteries have long been considered strong candidate solutions owing to their small spatial footprint, mechanical simplicity and flexibility in siting. However, the barrier to widespread adoption of batteries is their high cost. Here we describe a lithiumantimony-lead liquid metal battery that potentially meets the performance specifications for stationary energy storage applications. This LijjSb-Pb battery comprises a liquid lithium negative electrode, a molten salt electrolyte, and a liquid antimony-lead alloy positive electrode, which self-segregate by density into three distinct layers owing to the immiscibility of the contiguous salt and metal phases. The all-liquid construction confers the advantages of higher current density, longer cycle life and simpler manufacturing of large-scale storage systems (because no membranes or separators are involved) relative to those of conventional batteries 5,6 . At charge-discharge current densities of 275 milliamperes per square centimetre, the cells cycled at 450 degrees Celsius with 98 per cent Coulombic efficiency and 73 per cent round-trip energy efficiency. To provide evidence of their high power capability, the cells were discharged and charged at current densities as high as 1,000 milliamperes per square centimetre. Measured capacity loss after operation for 1,800 hours (more than 450 charge-discharge cycles at 100 per cent depth of discharge) projects retention of over 85 per cent of initial capacity after ten years of daily cycling. Our results demonstrate that alloying a high-meltingpoint, high-voltage metal (antimony) with a low-melting-point, lowcost metal (lead) advantageously decreases the operating temperature while maintaining a high cell voltage. Apart from the fact that this finding puts us on a desirable cost trajectory, this approach may well be more broadly applicable to other battery chemistries.
Among metalloids and semi-metals, Sb stands as a promising positive-electrode candidate for its low cost (US$1.23 mol -1 ) and relatively high cell voltage when coupled with an alkali or alkaline-earth negative electrode 5 . In previous work 6 , we demonstrated the performance of a MgjjSb liquid metal battery at current densities ranging from 50 to 200 mA cm -2 , achieving a round-trip energy efficiency of up to 69%. However, the high melting points of Mg (T m 5 650 uC) and Sb (T m 5 631 uC) require the cell to operate near 700 uC. A high operating temperature is undesirable because it results in higher rates of corrosion and detracts from overall storage efficiency, which ultimately increases cost of ownership. These potential limitations, in conjunction with an estimated electrode materials cost of US$375 kWh 21 and an average cell voltage of 0.21 V (measured under galvanostatic discharge at 200 mA cm -2 ), render MgjjSb cells impractical for commercial applications 5 .
With an average cell voltage of 0.92 V, the LijjSb combination is an appealing alternative 7 . Moreover, Li melts at 180 uC and exhibits low solubility in lithium halide melts, which results in lower self-discharge current and, hence, higher energy efficiency especially when compared with sodium alternatives 8 . Despite these attractive properties, the high melting point of Sb sets the operating temperature of the LijjSb cell at almost 500 uC above the melting point of Li. Although alloying Sb with another metal can be an effective strategy to lower the melting point of the positive electrode, this is generally accompanied by an undesirable decrease in cell voltage, as observed in the MgjjSn-Sb (ref. 9) and NajjBi-Sb systems (ref. 10).
Here we report that, with the use of a Li negative electrode, the addition of Pb to Sb maintains a cell voltage almost as great as that for pure Sb while substantially reducing the melting temperature (eutectic composition, Sb-Pb 18:82 mol%; melting temperature, T m 5 253 uC; ref. 11).  The corresponding equilibrium voltages of the binary LijjSb cell 12 and the binary LijjPb cell 13 are also shown for comparison. At the chosen experimental temperature (450 uC), Sb-Pb alloys and pure Pb are liquid, whereas Sb is solid. As shown in Fig. 1, the LijjSb (solid) chemistry has the highest cell voltage, at 0.92 V, and the LijjPb chemistry exhibits the lowest cell voltage, just under 0.6 V. We note that even at high dilution (up to 82 mol% Pb in Sb), the LijjSb-Pb systems operate at cell voltages very near LijjSb levels (only about 0.05 V lower), indicating that the LijjSb-Pb electrode potential is determined primarily by the Li-Sb interaction. To reveal the predominant role of Sb in setting the potential, Fig. 1b was normalized to the concentration of Li relative to Sb. Figure 1b shows that all Li-Sb-Pb electrodes share a behaviour similar to that of the Li-Sb electrode: in each case, a region of near-constant, high potential is followed by a drop in potential at 75 mol% Li relative to Sb. This behaviour is suggestive of the LijjSb system in which Li x Sb (x , 3) compounds are formed and thereby generate a high electrode potential relative to pure Li. When the concentration of the alloy reaches 75 mol% Li in Sb, a low-potential Li 3 Sb phase is formed, and the cell voltage drops precipitously 7 .
Cell performance was demonstrated in 1.9 Ah theoretical capacity cells (3.16 cm 2 positive-electrode/electrolyte interfacial area) fitted with a Li negative electrode, an Sb-Pb positive electrode (30-70 mol%) and a LiF-LiCl-LiI molten salt electrolyte (20:50:30 mol%, T m 5 430 uC). Cells were assembled in the fully charged state in an Ar-filled glove box, placed inside a sealed test vessel and operated in a vertical tube furnace at 450 uC. When the temperature exceeded the melting point of the salt, the equilibrium cell voltage stabilized at ,1.0 V, consistent with titration results. At a stepped-potential of 1.2 V, the self-discharge current was measured to be 0.6 mA cm -2 , which is significantly lower than the value observed in Na systems 8 (20 mA cm -2 ). This is attributable to the lower solubility of Li in its molten halides 14 . A typical charge-discharge voltage profile and the usual performance metrics as functions of cycle index are shown in Fig. 2. At a high current density of 275 mA cm -2 , cells consistently achieved 93% of their theoretical capacity. The nominal discharge voltage was 0.73 V, which is more than three times higher than that of MgjjSb. On the basis of measured cell performance, the electrode materials costs are estimated to be US$68 kWh 21 , which is about one-fifth of the value for MgjjSb cells.
For stationary applications, long service lifetime is a critical factor. Liquid metal batteries are advantageous owing to the liquid electrodes and molten salt electrolyte, which avoid many of the common failure mechanisms associated with batteries fitted with solid-state electrodes, for example undesirable film formation at the electrode-electrolyte interface, or phase transformations that mechanically damage cell components 15 . Here we show the results of charge-discharge cycle testing of the LijjSb-Pb system (Fig. 2b). Over the duration of the test, the cells exhibited a Coulombic efficiency of 98% and a round-trip energy efficiency of 73%, maintaining 94% of the initial capacity after 450 cycles at full depth of discharge. The capacity fade rate decreased after the 100th cycle, whereupon the fade rate between the 100th and 450th cycles was 0.004% per cycle. This is equivalent to retention of over 85% of the initial capacity after ten years of daily cycling. Cross-sections of cells after 1,800 hours of operation did not exhibit any obvious corrosion of cell components (current collectors and walls).
The ability to operate at high current densities with minimal impact on cycle life is an asset for certain grid applications such as ancillary services. Here we show the capability of LijjSb-Pb cells with a LiF-LiCl-LiI salt electrolyte (20:50:30 mol%, T m 5 430 uC) operating at current densities as high as 1,000 mA cm -2 (Fig. 3) not only while discharging, but also while charging. The latter is especially useful in applications such as frequency regulation. At current densities as high as 500 mA cm -2 ,

LETTER RESEARCH
there is no significant decrease in the reversible capacity (87% of the theoretical value). Even at the highest current density (1,000 mA cm -2 ), the cell performed at 54% of its theoretical capacity. Most noteworthy about this last observation is the ability of the cell to act as a high-current load without incurring permanent damage: efficiency in this instance is subordinate to long-term electrode stability. This advantageous mix of features is attributable to the rare combination of the high conductivity of the molten salt electrolyte, ultrafast charge-transfer kinetics at the electrode-electrolyte interface between the liquid metal and molten salt, and fast mass transport within the liquid metal electrodes.
To demonstrate the scalability of the system, cells with a 62 Ah theoretical capacity (62 cm 2 ) were constructed and operated with performance similar to that achieved on the smaller scale (1.9 Ah). To optimize systems costs, a LiF-LiCl-LiBr eutectic electrolyte (22:31:47 mol%, T m 5 443 uC) and a positive-electrode Sb-Pb composition of 40:60 mol% were chosen. Operating at a temperature of 500 uC, the cells were cycled at 275 mA cm -2 . Figure 4 shows a typical charge-discharge voltage profile and the cell performance metrics over twenty cycles. With average values of Coulombic efficiency of 98% and a round-trip energy efficiency of 71%, negligible capacity fade was observed over the duration of the test. The nominal discharge voltage was measured to be ,0.69 V. On this basis, the electrode materials costs were estimated to be US$65 kWh 21 .
Alloying Sb with Pb has been identified as a way to achieve significant reductions in the melting point of the positive electrode as well as the cell operating temperature without an attendant decrease in cell voltage. This finding not only lowers the cost of LijjSb-Pb batteries, increasing their attractiveness for stationary applications, but also serves as an example of how to broaden the selection of positive-electrode materials for liquid metal batteries.
Online Content Methods, along with any additional Extended Data display items and Source Data, are available in the online version of the paper; references unique to these sections appear only in the online paper.

METHODS
For all experiments, high purity (.99.9%), ultradry-grade LiF, LiCl, LiBr and LiI salts (Alfa Aesar) were used in electrolytes. Salt mixtures were dried under vacuum at 80 uC for 8 h and 250 uC for 2 h to remove residual water, and were then melted under Ar gas at 600 uC. All experiments were performed under a high-purity Ar atmosphere.
For the evaluation of equilibrium voltage by coulometric titration, all alloy target compositions were prepared using an arc-melter (MAM1, Edmund Bühler GmbH). After arc-melting, alloys were placed into a small mullite basket, pre-melted and used as working electrodes. Electrical contact was established using a tungsten wire (1 mm diameter) immersed in the Sb-Pb alloy. Ag/AgCl served as the reference electrode (3 wt% AgCl in a LiCl-NaCl-KCl eutectic mixture) and was contained within a closed-end mullite tube with the end polished to a thin microporous layer. A 40:60 mol% Li-Al alloy was used as the counter-electrode. Electrochemical measurements were performed with an Autolab PGSTAT 302N potentiostat/galvanostat.
For cell testing, galvanostatic charge and discharge were performed using an Arbin BT2000. Alloys were pre-weighed and placed in cell containers. The salt mixtures were introduced and dried in situ under vacuum at 80 uC for 8 h and 250 uC for 4 h before setting the operating temperature for electrochemical testing. A cell schematic of the LijjSb-Pb liquid metal battery is shown in Extended Data Fig. 1.
All electrode cost estimations were performed using the following formula: C 5 S i P i m i /E, where C is the capital cost per unit of discharged electrical energy in US$ kWh 21 , P i is the specific bulk metal cost in US$ kg 21 , m i is the metal's mass in kg, and E is the discharged energy in kWh. For LijjSb-Pb cells, the value of discharged energy, E, was that measured during galvanostatic cycling at 275 mA cm -2 (Figs 2  and 4). For MgjjSb cells, the discharged energy, E, was that measured during galvanostatic cycling at 200 mA cm -2 (ref. 6). Average bulk metal prices of Li (US$61.7 kg -1 ), Mg (UD$2.7 kg -1 ), Pb (US$2.1 kg -1 ) and Sb (US$10.1 kg -1 ) were obtained from the MetalPrices Online Database (http://www.metalprices.com literature). Detailed cost calculations for the cells are presented in Extended Data Tables 1-3. A balance of system and salt costs is not included because the technology has yet to be fully developed on the commercial scale, and so there is no accurate basis for such estimation.