Nature | Letter
An aqueous, polymer-based redox-flow battery using non-corrosive, safe, and low-cost materials
- Journal name:
- Nature
- Volume:
- 527,
- Pages:
- 78–81
- Date published:
- DOI:
- doi:10.1038/nature15746
- Received
- Accepted
- Published online
For renewable energy sources such as solar, wind, and hydroelectric to be effectively used in the grid of the future, flexible and scalable energy-storage solutions are necessary to mitigate output fluctuations1. Redox-flow batteries (RFBs) were first built in the 1940s2 and are considered a promising large-scale energy-storage technology1, 3, 4. A limited number of redox-active materials4, 5, 6, 7, 8, 9, 10—mainly metal salts, corrosive halogens, and low-molar-mass organic compounds—have been investigated as active materials, and only a few membrane materials3, 5, 11, 12, 13, 14, such as Nafion, have been considered for RFBs. However, for systems that are intended for both domestic and large-scale use, safety and cost must be taken into account as well as energy density and capacity, particularly regarding long-term access to metal resources, which places limits on the lithium-ion-based and vanadium-based RFB development15, 16. Here we describe an affordable, safe, and scalable battery system, which uses organic polymers as the charge-storage material in combination with inexpensive dialysis membranes, which separate the anode and the cathode by the retention of the non-metallic, active (macro-molecular) species, and an aqueous sodium chloride solution as the electrolyte. This water- and polymer-based RFB has an energy density of 10 watt hours per litre, current densities of up to 100 milliamperes per square centimetre, and stable long-term cycling capability. The polymer-based RFB we present uses an environmentally benign sodium chloride solution and cheap, commercially available filter membranes instead of highly corrosive acid electrolytes and expensive membrane materials.
Subject terms:
At a glance
Figures
-
Figure 1: Working principle of a polymer-based RFB. a, Schematic representation of a polymer-based RFB consisting of an electrochemical cell (which determines the power density) and two electrolyte reservoirs (which determine the storage capacity). The anolyte and catholyte cycle are separated by a semipermeable size-exclusion membrane, which retains the redox-active macromolecules while allowing small salt ions to pass. During the charging/discharging process, a solution of the redox-active polymers P1 and P2 is continuously transported from the electrolyte reservoirs to the electrochemical cell, where the redox reactions take place. b, Fundamental electrode reactions of P1 (TEMPO radical) and P2 (viologen). Structural details of compounds shown in this figure are available as Supplementary Information.
-
Figure 2: Charge/discharge behaviour. A representative cell voltage profile of a pumped 5-cm2 test cell during constant-current cycling at 40 mA cm−2 with 10 ml of P1 and 15 ml of P2 solution (charge storage capacity adjusted to 10 A h l−1 in aqueous NaCl solution (2 mol l−1), 25 °C).
-
Figure 3: Electric performance of the polymer-based RFB cell. a, b, The capacity, coulombic efficiency, and energy efficiency of a pumped 5-cm2 test cell (10 ml of P1 and 15 ml of P2 aqueous NaCl solution (2 mol l−1); storage capacity adjusted to 10 A h l−1, 25 °C) as a function of discharging current density (a; charging at 40 mA cm−2) and charging current density (b; discharging at 40 mA cm−2).
-
Figure 4: Cycling stability of the polymer-based RFB. a, The long-term stability of the polymer-based electrolytes was studied by repeated charge/discharge cycling over 10,000 cycles at 20 mA cm−2 in an unpumped test cell. Inset, the open-circuit voltage of a polymer-based RFB as a function of the state of charge (static 5-cm2 test cell, P1 (storage capacity of 2 A h l−1) and P2 (storage capacity of 4 A h l−1) in aqueous NaCl solution (2 mol l−1), 25 °C). b, A pumped cell was repeatedly cycled at 40 mA cm−2 (10 ml of P1 and 15 ml of P2 in aqueous NaCl solution (2 mol l−1); storage capacity adjusted to 10 A h l−1, 25 °C).
-
Extended Data Fig. 1: Schematic representation of the synthesis of redox-active polymers. The cathode material P1 and anode material P2 were prepared by free radical polymerization and subsequent polymer-analogous oxidation and functionalization, respectively. ABCVA, 4,4′-azobis(4-cyanovaleric acid); AIBN, azobisisobutyronitrile; DMSO, dimethyl sulfoxide.
-
Extended Data Fig. 2: Rheogram of redox-active polymers. The flow behaviour of aqueous solutions of P1 and P2 was studied at 20 °C under continuous shear in sodium chloride solution (1 mol l−1) using a double-gap measuring system in a rotational rheometer. The concentrations of the polymers correspond to a charge-storage capacity of about 10 A h l−1.
-
Extended Data Fig. 3: Test set-up for a polymer-based RFB. a, Photograph of a laboratory set-up (5-cm2 test cell, peristaltic pump, and electrolyte reservoirs) used for charging/discharging experiments. b, Exploded-view drawing of the 5-cm2 test cell. c, Colour change of the polymer solutions upon charging and discharging.
-
Extended Data Fig. 4: Crossover studies on dialysis membrane. a, Time-dependent NaCl concentration of a chamber filled with deionized water that is separated from a NaCl feed solution (1 mol l−1) by a dialysis membrane. Salt permeability was determined to be Ps = (9.3 ± 0.1) × 10−5 cm s−1. b, Time-dependent P1 concentration of an RFB test cell compartment filled with NaCl solution (2 mol l−1) that is separated from a P1 feed solution (60 mg ml−1) by a membrane. We determined the maximum polymer permeability from the linear part of the diffusion graph to be Ppolymer,max = (3.2 ± 0.3) × 10−7 cm s−1 (diffusion coefficient Dpolymer,max = (1.3 ± 0.1) × 10−7 cm2 min−1). The minimum membrane selectivity Smin = 290. c, Time-dependent change in P2 concentration of an RFB test cell compartment filled with NaCl solution (2 mol l−1) that is separated from a P2 feed solution (40 mg ml−1) by a dialysis membrane. Ppolymer,max = (3.3 ± 0.4) × 10−8 cm s−1; Dpolymer,max = (1.4 ± 0.2) × 10−8 cm2 min−1; Smin = 2,830. All crossover experiments were conducted with a cellulose-based dialysis membrane (MWCO = 6,000 g mol−1) at 25 °C. In b and c, the slopes of the fit lines correspond to
in
(see Methods). d, The number-weighted distributions of hydrodynamic radii (
Rh
n,app) of P1 and P2 determined by dynamic light scattering reveal mean radii of approximately 2 nm (5 g l−1 in 0.1 mol l−1 NaCl solution). Inset, a comparison of the intensity- and number-weighted distributions of P1 shows the presence of aggregates (4.6 nm and 184 nm). -
Extended Data Fig. 5: Cyclic voltammogram of electrolytes after 10,000 cycles. a, b, The cyclic voltammogram (in water with 0.1 mol l−1 NaCl; scan rate of 200 mV s−1) of samples taken from the anolyte (a) and catholyte (b) after repeated charging/discharging; solid lines and dashed lines correspond to the reductive and oxidative range, respectively, of the cyclic voltammogram.
-
Extended Data Fig. 6: Electrochemical analysis of P1 and P2. a, Cyclic voltammogram of the oxidation process of P1 (2.5 × 10−3 mol l−1 in water with 0.1 mol l−1 NaCl; scan rate of 200 mV s−1). b, Cyclic voltammogram of the first reduction process of P2 (5.2 × 10−3 mol l−1 in water with 0.1 mol l−1 NaCl; scan rate of 200 mV s−1). c, Ultraviolet–visual spectroelectrochemistry of P2 at different applied potentials (as indicated) during consecutive reduction and after subsequent re-oxidation (10−4 mol l−1 in water with 0.1 mol l−1 NaCl). The single-reduced species (Viol+•) is visible at −450 mV (orange line) with the formation of three distinct bands at 365 nm, 530 nm, and 900 nm, which disappear upon further reduction towards the double-reduced species (Viol0). The shapes and positions of the emerging bands strongly suggest the formation of radical cation dimers27. Re-oxidation at 200 mV (dotted line) restores the initial spectrum.
-
Extended Data Fig. 7: Toxicity tests of the redox-active polymers. a, The viability of L929 mouse cells was tested in the presence of redox-active compounds according to ISO10993-5. Cell viability below 70% is considered indicative of cytotoxicity. The negative control was standardized as 0% of metabolism inhibition and referred as 100% viability. Two vanadium salts in redox states commonly found in vanadium-based RFBs and two widely used cationic polymers were used as a reference. Poly(l-lysine), PLL, is a commercial food preservative and branched poly(ethylene imine), bPEI, is used in the paper-making industry and as a flocculating agent. Although P1 and P2 show cytotoxic effects at concentrations >50 µg ml−1—with P1 being less toxic than P2—the vanadium salts and the cationic polymers reveal cytotoxic effects at lower concentrations (VOSO4 > 5 µg ml−1, VCl3 > 10 µg ml−1, bPEI > 5 µg ml−1, PLL > 25 µg ml−1). Data are expressed as mean values and error bars represent the standard deviation of three determinations. b, We quantified the cell-membrane damaging properties of the polymers by analysing the haemoglobin release from erythrocytes (indicated by the numbers associated with each bar). Data are expressed as mean values and error bars represent the standard deviation of triplicates of three different blood samples per concentration. Because a haemoglobin release (haemolysis rate) less than 2% is considered non-haemolytic, P1 and P2 as well as the reference compounds show no membrane damaging behaviour. Hence, the cell toxic effects do not originate from damage to the cell membrane, but from reactions within the cell. Because the cell uptake via diffusive processes of polymers is hindered in comparison to ‘small’ inorganic ions, P1 and P2 possess lower cytotoxicity. These tests provide some insight into the toxicity, but long-term ecotoxicity tests and animal testing are required to fully evaluate the impact of the redox-active polymers on wildlife and plants.
-
Extended Data Fig. 8: RDE measurements of P1 and analysis. a, Voltammograms of P1 (2.5 × 10−3 mol l−1 in water with 0.1 mol l−1 NaCl; scan rate of 5 mV s−1) at different rotation speeds (as indicated) from 400 r.p.m. to 3,600 r.p.m. (The arrow indicates the direction of potential scanning). b, Levich plot from the obtained limiting currents; application of Levich equation yields a diffusion coefficient D = (7.0 ± 0.5) × 10−8 cm2 s−1. c, Koutecký–Levich plot for different overpotentials η (as indicated) yielding the mass-transfer-independent current ik (as ω−1/2
0, ik = i). d, Tafel plot yielding k0 = (4.5 ± 0.1) × 10−4 cm s−1 and α = 0.68 ± 0.03 (see Methods). -
Extended Data Fig. 9: RDE measurements of P2 and analysis. a, Voltammograms of P2 (5.2 × 10−3 mol l−1 in water with 0.1 mol l−1 NaCl; scan rate 5 mV s−1) at different rotation speeds (as indicated) from 100 r.p.m. to 4,900 r.p.m.; substantial changes of the limiting current, necessary for reasonable analysis, are observed only at rotation speeds >1,200 r.p.m. (The arrow indicates the direction of potential scanning). b, Levich plot for currents between the first and second steps of the two-step process; the Levich equation was applied only for high rotation rates, that is, in the region of substantial changes of the limiting current, yielding a diffusion coefficient D = (7.6 ± 0.9) × 10−7 cm2 s−1. The fit curve and its slope correspond to the Levich equation (see Methods). c, Plot of i−1 versus ω−1/2 for high negative overpotentials η (as indicated) yielding 1/ik (as ω−1/2
0, 1/ik = 1/i). d, Plot of log|1/ik| versus η1 (overpotential with respect to the first step) yielding −log|2FAk0c0| (log|1/ik| for η1 = 0), which allows us to determine k0 = (9 ± 2) × 10−5 cm s−1. The error bars represent error that originates from the linear regression analysis of c.