Suppressing corrosion in primary aluminum–air batteries via oil displacement

See allHide authors and affiliations

Science  09 Nov 2018:
Vol. 362, Issue 6415, pp. 658-661
DOI: 10.1126/science.aat9149

Oil when not in use

For primary or nonrechargeable batteries, the overall energy density will be limited by any discharge or open-circuit corrosion that occurs during storage. For batteries based on aluminum and air, this longstanding problem has prevented their widespread use and has been challenging to overcome. Hopkins et al. used commercially available components to construct aluminum-air batteries. During standby periods, the electrolyte in the batteries was replaced with oil to protect the electrodes from corrosion, thus preventing energy loss.

Science, this issue p. 658


Primary aluminum–air batteries boast high theoretical energy densities, but negative electrode corrosion irreversibly limits their shelf life. Most corrosion mitigation methods are insufficient or compromise power and energy density. We suppressed open-circuit corrosion by displacing electrolyte from the electrode surface with a nonconducting oil during battery standby. High power and energy density are enabled by displacing the oil with electrolyte for battery discharge. The underwater-oleophobic wetting properties of the designed cell surfaces allow for reversible oil displacement. We demonstrate this method in an aluminum–air cell that achieves a 420% increase in usable energy density and 99.99% reduction in corrosion, which lowers self-discharge to a rate of 0.02% a month and enables system energy densities of 700 watt-hours per liter and 900 watt-hours per kilogram.

Single-use and mechanically rechargeable aluminum–air (Al–air) batteries are low-cost options for applications that require ultrahigh energy densities that electrically rechargeable batteries have yet to achieve (1, 2). Such applications include long-range drones (3), off-grid power supplies (4, 5), and range extenders for electric vehicles (6, 7). A long-standing barrier to their increased adoption, however, has been severe open-circuit corrosion that curtails their shelf life (8, 9). Despite recent advances in corrosion inhibition (10), Al–air batteries can irreversibly self-discharge more than 80% a month depending on battery design (11), while rechargeable lithium-ion batteries self-discharge only 5% a month (12).

Methods to mitigate corrosion in Al–air batteries typically require a substantial trade-off between power and energy density and open-circuit corrosion. Mitigation methods include negative electrode alloying (2), electrolyte additives (2), gel electrolytes (13), nonaqueous electrolytes (14), and draining the electrolyte after shutdown (4, 15, 16). Alloying combined with aqueous electrolyte additives yields high power and energy density but also high corrosion currents near 1 to 10 mA cmgeo−2 (geometric area) (1, 2, 10). Alloys generally use greater than 99.99 weight % (wt %) purity Al combined with a variety of elements such as Zn, In, Mg, Ga, Sn, Bi, B, and Mn (2). Studied electrolyte additives include ZnO, SnO32–, In(OH)3, BiO33–, Ga(OH)4, MnO42–, Cl, NO3, SO42–, and SnO32– (2). Conversely, gel and nonaqueous electrolytes using polyacrylic acid and ionic liquids achieve lower corrosion rates but only fractions of the power and energy density achieved with more conductive aqueous electrolytes (13, 14). Ionic electrolytes attain the lowest reported open-circuit corrosion currents, near 0.002 mA cmgeo−2 (14). Draining the electrolyte after shutdown can result in considerable electrode corrosion and hazardous clogging of the hydraulic system of an Al–air battery pack (15, 16). After draining, remaining electrolyte clings to the hydrophilic Al electrodes, induces corrosion, dries up, and leaves crusts of by-products and electrolyte solutes (15, 16). Rinsing the Al electrode with water or a pH-neutralizing agent after electrolyte draining improves restart performance but can significantly reduce system energy density if the rinsing system is contained in the battery pack (4, 5, 16). Such rinsing systems require additional tanks for water and electrolyte along with a neutralizing agent that is consumed with every shutdown (4, 5, 16). If the trade-off between power and energy density and open-circuit corrosion could be sufficiently mitigated, the use of Al–air batteries would extend beyond current niche applications.

We overcame this performance trade-off by redesigning the conventional flowing electrolyte metal–air battery (Fig. 1A) (5, 16, 17) with a separator that displays underwater oleophobicity (Fig. 1, B and C) (18). This separator allows for the reversible displacement of electrolyte from the negative electrode surface with an oil when the battery is on standby. The nonconducting displacing oil dramatically reduces the diffusion rate of electrolyte to the negative electrode surface, enabling ultralow open-circuit corrosion. We achieved high power and energy density by displacing the oil with electrolyte for battery discharge. Such a method could be used to suppress open-circuit corrosion in primary Al–air cells and potentially other batteries (8, 9).

Fig. 1 Oil displacement method.

(A) Schematic of the conventional flowing electrolyte metal–air battery (16, 17). (B) Schematic of the constructed oil displacement system for a flowing electrolyte metal–air battery. Electrolyte is continuously pumped during operation. When not in use, oil is pumped to displace the corrosive electrolyte for a specified duration. (C) Schematic of the constructed oil displacement system when not in use, with a magnified view of the interfaces of the metal electrode and separator. (D) Voltage versus time for on-off cycling of a primary Al–air battery with (A) the conventional cell design and [(B) and (C)] the constructed cell design. Current densities of 150 mA cmgeo−2 were drawn for 5 min with 24- or 72-hour pauses during which no current was drawn. The conventional cell stopped operating at the start of day 3, yielding an energy density of 0.40 ± 0.07 Wh gAl−1. By contrast, the constructed cell lasted more than 24 days, yielding an energy density of 2.08 ± 0.07 Wh gAl−1.

To demonstrate this concept, we constructed a primary flowing-electrolyte Al–air battery (fig. S1). The cathode was a commercially available air electrode (Quantum Sphere). The air electrode consisted of carbon particles and nano-manganese catalyst pressed on a nickel mesh layered with an air-breathing, waterproof expanded polytetrafluoroethylene (ePTFE) membrane that enables a reduction reaction with oxygen from the atmosphere (Eq. 1) (19). The anode was a 0.25-mm-thick nonporous Al foil (99.999 wt % purity) (Sigma-Aldrich) that oxidizes during discharge (Eq. 2). The selected aqueous electrolyte was the conventionally used, highly conductive 4 M sodium hydroxide (NaOH) (98 wt % purity) paired with a common corrosion inhibitor, 0.05 M sodium stannate (Na2SnO3) (95 wt % purity) (20).O2 + 2H2O + 4e → 4OH Ec = 0.3 V versus Hg/HgO(1)Al + 4OH → Al(OH)4 + 3e Ea = –2.4V versus Hg/HgO(2)We chose low–surface tension displacing oils and hydrophilic separators for testing in the constructed cell because such systems are likely to display underwater oleophobicity and be resistant to oil fouling. This conclusion can be drawn from a two-phase Young’s equation (Eq. 3) relating the wetting properties of fluids in air to the wetting properties of those fluids submerged in a different fluid (21).Embedded Image(3)For this application, the variable θoe is the Young contact angle of an oil droplet surrounded by electrolyte on the negative electrode or separator surface. The variables θo–air and θe–air are the Young contact angles of an oil and electrolyte droplet surrounded by air on either the negative electrode or separator surface, respectively. The surface tension of the oil and electrolyte are respectively γo–air and γe–air, with an oil-electrolyte interfacial tension expressed as γoe. Equation 3 suggests that to achieve underwater oleophobicity, where θoe is greater than 90°, the numerator in the expression right of the equals sign must be negative. The more negative the numerator, the larger the underwater contact angle. To achieve a negative numerator, a low–surface tension oil (γo–air) can be selected along with a hydrophilic negative electrode and separator (θe–air < 90°). We therefore selected two low–surface tension candidate oils, silicone and perfluoropolyether (PFPE) oil (22), that have large working temperatures (–40° to 100°C), act as ionic insulators to inhibit corrosion, and cannot saponify in high-pH solutions because they contain no triglycerides. We also chose two hydrophilic separators, Celgard 5550 and a PTFE separator (Advantec MFS) commercially treated with an additive to be hydrophilic (11), that display high working temperatures and chemical compatibility with the electrolyte.

We identified appropriate separator pore sizes using the Young-Laplace equation to ensure that the pores suppressed the passage of oil through the separator at appropriate pumping pressures. The Young-Laplace equation, ΔP = 4γoe/D, estimates the pressure differential ΔP that is required for oil to leak through a circular pore with a diameter D given an interfacial tension between the oil and electrolyte of γoe. For example, the measured interfacial tension between the electrolyte and PFPE oil was 61.6 ± 0.1 mN m−1 (11). Commercially available pore diameters for hydrophilic PTFE separators range from 0.1 to 1 μm. We selected a 1-μm pore size to minimize ionic resistance across the separator. If we assume severe thermal expansion of 100% owing to numerous heating cycles, 2-μm pore diameters would allow for high pumping pressures of 123 kPa (23).

The oil displacement method is enabled by using an oil that displays underwater oleophobicity on both the negative electrode and separator surfaces. To demonstrate this principle, we measured advancing and receding underwater contact angles of silicone oil on the Celgard separator and on the Al electrode and of PFPE oil on the PTFE separator and on the Al electrode (fig. S5). Both the silicone and PFPE oil displayed underwater oleophobicity on the Al electrode (Fig. 2, A and B, top left), but the silicone oil wetted the Celgard, whereas the PFPE oil beaded up on the hydrophilic PTFE separator (Fig. 2, A and B, bottom left). We then performed on-off cycling with Al–air cells using the respective oil-separator combinations (Fig. 2, A and B, right). Current densities of 150 mA cmgeo−2 were drawn for 25 min with 1-hour pauses during which no current was drawn (11). At the start of each pause, the displacing oils were pumped into the batteries for 50 s, requiring less than 50 mW of power. Before the next discharge, the electrolyte was pumped into the batteries for 50 s. After the silicone oil was pumped in, subsequent discharge voltages were negative, suggesting that the silicone oil fouled the Celgard separator (Fig. 2A, right). The total energy density achieved with the silicone-Celgard combination was 0.66 ± 0.08 watt-hour (Wh) per gram Al (gAl −1) (11). By contrast, the PFPE-PTFE combination showed positive voltages after each pause (Fig. 2B, right). The total energy density achieved with the PFPE-PTFE combination was 3.55 ± 0.08 Wh gAl−1 (11), which is comparable with the energy density achieved by discharging the battery without pauses and without introducing oil, 3.54 ± 0.08 Wh gAl−1 (fig. S6) (11). These energy-density results imply that the Al electrode and hydrophilic PTFE separator resisted fouling by the PFPE oil.

Fig. 2 Selecting a displacing oil and separator.

(A) Contact angle schematics of silicone oil on the Al electrode (top left) and on a Celgard separator (bottom left). (Right) Voltage versus time for on-off cycling by using the silicone oil and Celgard separator. Current densities of 150 mA cmgeo−2 were drawn for 25 min, with 1-hour pauses during which no current was drawn. At the start of each pause, silicone oil was pumped into the battery for 50 s. Before the following discharge, electrolyte was pumped into the battery for 50 s. (Inset) Schematic of the Al electrode and separator surfaces. The silicone oil fouls the Celgard separator but not the Al electrode. (B) Corresponding experiments presented in (A), using a PFPE oil and hydrophilic PTFE separator. (Inset) A fouling-resistant hydrophilic PTFE separator and Al electrode. The Al–air battery using the silicone-Celgard combination yielded a negative voltage after the first 1-hour pause, resulting in an energy density of 0.66 ± 0.08 Wh gAl−1, whereas the Al–air battery using the PFPE-PTFE combination reached full discharge capacity, yielding an energy density of 3.55 ± 0.08 Wh gAl−1. Contact-angle photos can be found in fig. S5.

Perturbations and electrolyte pumping trigger the displacement of oil by electrolyte. The Al electrode and separator’s underwater oleophobicity (fig. S5) promote the rupturing of thin oil films on their surfaces, as suggested by thin film rupture theory (24). The perturbations and electrolyte pumping induce flow instabilities (24, 25) that displace the oil from the electrode and separator surfaces. For displacement of electrolyte by oil, the bulk electrolyte is displaced by the flowing oil that thins the remaining electrolyte films, which are eventually consumed by corrosion on the Al electrode.

With the selected PFPE displacing oil and hydrophilic PTFE separator, we showed that the constructed cell (Fig. 1, B and C) achieved a greater usable energy density than that achieved with a conventional cell (Fig. 1A) by drawing current densities of 150 mA cmgeo−2 for 5 min with 24- or 72-hour pauses in between (Fig. 1D) (11). During pauses for the conventional cell, the electrolyte remained in the interelectrode gap, with corrosion currents of 1 to 10 mA cmgeo−2 (figs. S2 and S3) (11). For the constructed cell with the oil displacement system, PFPE oil was pumped into the cell for 50 s at the start of each pause to displace the corrosive electrolyte. Once the oil displaced the electrolyte, corrosion currents dropped to 0.0001 to 0.0011 mA cmgeo−2 (figs. S2 and S4) (11), which are below those achieved by state-of-the-art ionic electrolytes (14). Before each discharge, the electrolyte was pumped back into the cell for 50 s. At the start of day 3, the conventional cell yielded a negative voltage when 150 mA cmgeo−2 were drawn. The Al electrode was corroded through its thickness in the conventional cell, which was monitored through the transparent battery casing. The resulting energy density of the conventional cell was 0.40 ± 0.07 Wh gAl−1 (11). By contrast, the constructed cell using the oil displacement system lasted more than 24 days and yielded an energy density of 2.08 ± 0.07 Wh gAl−1 (11). Commercial Al–air anodes are relatively thick, 0.5 cm (26), and nonporous to mitigate corrosion. They sustain minimal loss in power as electrode thickness increases because no ionic diffusion occurs within the solid Al electrode during discharge (27). Consequently, a scaled-up 0.5-cm-thick Al electrode would achieve a self-discharge rate near 0.02 ± 0.02% a month, which could be tuned based on oil film thickness (11).

The performance of the constructed cell with the oil displacement system was comparable with the performance of Al–air cells in the literature. The open-circuit potential of the cell with and without the PTFE separator was 1.8 V as found in the literature (20). At 150 mA cmgeo−2, the coulombic efficiency with and without the separator was 91%, which also matched literature values (20). We measured the voltage (Fig. 3A and fig. S7) and capacity (Fig. 3B and fig. S8) as a function of current density. The prototype with the hydrophilic PTFE separator achieved a peak power density of 300 ± 17 mW cmgeo−2 (Fig. 3A and fig. S9) and energy density of 3.73 ± 0.08 Wh gAl −1 (fig. S10). Reported peak powers of flowing-electrolyte Al–air batteries with 2- to 3-mm interelectrode gaps range from 350 to 620 mW cmgeo−2 and obtain peak energy densities near 4.3 Wh gAl −1 (1, 17). Al–air batteries with gaps near 10 mm, however, achieve peak powers as low as 75 mW cmgeo−2 (28). We attribute our values to the PTFE separator and to the 7-mm interelectrode gap of the constructed prototype. Without the separator, the prototype reached a peak power of 350 ± 17 mW cmgeo−2 (fig. S9). The separator, however, minimally affected power and energy density for energy-efficient current densities below 250 mA cmgeo−2 (fig. S10). For example, at 150 mA cmgeo−2, the power and energy density achieved with the separator was 192 ± 5 mW cmgeo−2 (Fig. 3A) and 3.54 ± 0.08 Wh gAl −1 (fig. S6). Excluding the separator yielded 199 ± 4 mW cmgeo−2 (fig. S9) and 3.75 ± 0.08 Wh gAl−1 (fig. S6). The large 7-mm gap allowed for appropriate separator sealing and easy assembly and disassembly of the prototype cell with bolts (fig. S1). Production-level sealing methods such as laser or ultrasonic welding would enable smaller interelectrode gap sizes.

Fig. 3 Constructed-cell performance and pack-level characteristics.

(A) Voltage and power density versus current density of the constructed primary aqueous Al–air battery using the hydrophilic PTFE separator. (B) Voltage versus capacity of the same battery used in (A) for discharging currents of 50, 100, 150, 200, and 250 mA cmgeo−2. (C) Estimated mass distribution of an Al–air battery pack (1) with an oil displacement system. The oil accounted for 13% of the total mass of the system. (D) Usable system energy density versus the number of startup/shutdown cycles that occur before complete discharge.

Oil accounted for 13% of the total mass of an Al–air battery pack with an oil displacement system (Fig. 3C). Al–air electric vehicle packs achieve peak powers of 34.8 kW and maximum system energy densities of 936 Wh liter−1 and 1300 Wh kg−1, with total volumes and masses of 371 liters and 267 kg (1). Calculations include practical pack components such as an electrolyte pump, battery casing, and electrolyte filtration system. The energy density values assume that water is refilled four times before complete discharge and that discharge is continuous. To enable energy-efficient startup/shutdown capability, we added 20 liters of displacing oil to the pack model (1). This additional oil resulted in an increase of 36 kg of mass, assuming PFPE oil was used and represented 13% of the total pack mass.

We calculated scaled-up corrosion losses per shutdown cycle for an Al–air battery pack with an oil displacement system using data from the 24-day experiment (Fig. 1D). We attribute the reduction in capacity observed in the 24-day experiment with respect to the capacity achieved by continuous discharge (fig. S6) to corrosion associated with the thin layer of electrolyte that remained on the Al electrode after bulk electrolyte displacement. We observed minimal change in energy density when pauses were 1 hour (Fig. 2B) and minimal changes in voltage between pauses that lasted 24 and 72 hours (Fig. 1D). These observations suggest that little corrosion occurred after 1 hour and that remaining electrolyte layers were consumed by corrosion after ~24 hours. We therefore assumed equal volumes of remaining electrolyte on the negative electrode after each discharge of the 24-day experiment, yielding a loss of 14 gAl mgeo−2 per pause due to corrosion. If 150 mA cmgeo−2 is the normal operating current density, an approximate voltage of 1.22 V will be maintained (Fig. 1D) with a capacity of 2.697 Ah gAl−1 (Fig. 3B). Assuming operation at the stated voltage and capacity, system-level corrosion yields a loss of 267 Wh per shutdown with 5.8 mgeo2 of Al (11). We estimated system-level pumping losses to be 2% of operating power at 150 mA cm−2 (17), with a displacement time of 10 min, yielding a loss of 40 Wh per startup/shutdown cycle. The large difference in corrosion and pumping losses suggests that there is room for energy-density improvement by pumping optimization (29). The corrosion and pumping losses per startup/shutdown cycle enable the calculation of usable system energy density as a function of the number of startup/shutdown cycles before complete discharge (Fig. 3D).

We estimated maximum usable pack-level energy densities of 700 Wh litersys−1 (system volume) and 900 Wh kgsys−1 (system mass) (Fig. 3D) with a standard operational power of 11 kW. To calculate a lower-bound energy density limit, we assumed that each startup/shutdown cycle yields corrosion damage comparable with a 24-hour shutdown. The usable system energy density (Esys) is a function of the energy stored in the Al (EAl), the number of startup/shutdown cycles (n), the losses due to corrosion and pumping per startup/shutdown cycle (closs and ploss, respectively), and the system volume or mass that includes the added mass and volume of the oil displacement system (vsys and msys, respectively), resulting in Esys = [EAln(closs + ploss)]/vsys for system volumetric energy density. Using the constructed prototype’s power at a current density of 150 mA cmgeo−2 yields a pack power of 11 kW. Such powers are sufficient for many applications, including industrial tools (30), off-grid power units (15), and vehicular propulsion (31). Applications with few startup/shutdown cycles would have energy densities near 700 Wh litersys−1 and 900 Wh kgsys−1. By contrast, applications that require hundreds of startup/shutdown cycles before complete discharge would achieve energy densities near 530 Wh litersys−1 and 680 Wh kgsys−1 (11). These calculations and data suggest that an Al–air pack with an oil displacement system achieves high system energy densities, power densities that are comparable with aqueous flowing electrolyte Al–air systems, and open-circuit corrosion currents that are lower than those attained with nonaqueous Al–air cells.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S10

References (3236)

Data File S1

References and Notes

  1. Materials and methods are available as supplementary materials.
Acknowledgments: Funding: Research was supported by the Massachusetts Institute of Technology (MIT) Lincoln Laboratory (awards 7000308296, 7000344422, and 7000401832). Author contributions: B.J.H. and D.P.H. conceptualized the oil displacement method. B.J.H., Y.S-H., and D.P.H. developed the methodology to test and characterize the performance of the oil displacement method. B.J.H. performed all experimental investigation, formal analysis, experimental validation, data curation, and software programing to process and present collected data. B.J.H. designed, built, and tested all battery systems presented in the work. Y.S-H. provided electrochemical instrumentation. B.J.H. prepared the initial manuscript and figures that D.P.H. and Y.S-H. reviewed and edited. D.P.H. supervised and acquired funding for the research. Competing interests: B.J.H. and D.P.H. are inventors on Patent Cooperation Treaty serial no. PCT/US2017/020093 covering the oil displacement method. Data and material availability: All data are available in the manuscript or the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article