Tafel Plots from Experiment and First-Principles Theory - American

Jan 30, 2013 - in a bulk electrolysis cell with those obtained by first-principles theory is semiquantitative. The kinetic ... Despite their great pro...
0 downloads 0 Views 842KB Size
Letter pubs.acs.org/JPCL

Li−O2 Kinetic Overpotentials: Tafel Plots from Experiment and FirstPrinciples Theory V. Viswanathan,† J. K. Nørskov,†,‡ A. Speidel,§ R. Scheffler,§ S. Gowda,∥ and A. C. Luntz*,‡,∥ †

Department of Chemical Engineering, Stanford University, Stanford, California 94305, United States SUNCAT, SLAC National Accelerator Laboratory, Menlo Park, California 94025, United States § Volkswagen Group, Inc., Belmont, California 94002, United States ∥ Almaden Research Center, IBM Research, 650 Harry Road, San Jose, California 95120, United States ‡

S Supporting Information *

ABSTRACT: We report the current dependence of the fundamental kinetic overpotentials for Li−O2 discharge and charge (Tafel plots) that define the optimal cycle efficiency in a Li-air battery. Comparison of the unusual experimental Tafel plots obtained in a bulk electrolysis cell with those obtained by first-principles theory is semiquantitative. The kinetic overpotentials for any practical current density are very small, considerably less than polarization losses due to iR drops from the cell impedance in Li−O2 batteries. If only the kinetic overpotentials were present, then a discharge−charge voltaic cycle efficiency of ∼85% should be possible at ∼10 mA/cm2 superficial current density in a battery of ∼0.1 m2 total cathode area. We therefore suggest that minimizing the cell impedance is a more important problem than minimizing the kinetic overpotentials to develop higher current Li-air batteries. SECTION: Energy Conversion and Storage; Energy and Charge Transport

O

output potential or sudden death after some discharge capacity (see Figure S1 in the Supporting Information). Figure 1 shows a plot of Udis during discharge versus the discharge current i from typical Swagelok-type Li−O2 cells that have been used by us and others in many previous Li−O2 studies. (See the SI for details of the cell design, electrolytes, and operating parameters used in Figure 1.) For these cells, O2 consumption forming Li2O2 is the dominant electrochemical process in discharge and O2 evolution is the dominant (but not exclusive) process during charge.5 This Figure shows results using two types of high surface area (∼0.1 m2) carbon as cathodes, XC72, and AvCarb P50 carbon paper. Thus, even at 10 mA/cm2 superficial current density through the cell the current density normalized to the total surface area of the cathode (assuming it is all electrochemically active) is only ∼10 μA/cm2. Only at the lowest superficial current densities (≤0.1 mA/cm2) is the output potential plateau Udis ∝ log(i) as anticipated for typical Tafel electrochemical kinetic behavior. At higher i, Udis is simply linear with i. This linear dependence is characteristic of an iR drop due to the cell impedance, and this dominates the potential loss rather than the fundamental electrochemical kinetics at the electrodes. From the slopes in Figure 1, R ≈ 40Ω (XC72) and 80Ω (P50), in good agreement with R measured by electrochemical impedance spectroscopy for cells at open

ver the past few years, research into nonaqueous Li-air (or Li−O2) batteries has been extremely active because developing a successful Li-air battery could give a safe and costeffective secondary battery with a much higher specific energy than Li-ion batteries. The net electrochemical reaction in a nonaqueous Li-air battery is 2Li + O2 ⇄ Li2O2, with battery discharge described by the forward direction and charge described by the reverse direction. Despite their great promise, there are significant challenges to developing practical Li-air batteries,1,2 for example, electrolyte and cathode stability3−5 and charge transport through the insulating Li2O2.6,7 However, even if all practical challenges can successfully be overcome, the discharge−charge cycle efficiency must ultimately be limited by the kinetic overpotentials η = |U − U0| of the fundamental surface electrochemistry itself, where U is the potential across the two electrodes and U0 is the equilibrium potential. It is therefore essential to understand the limiting overpotentials achievable in Li-air batteries and to see if this fundamental limit is even capable of supporting a high energy and high power battery. This is best described in terms of Tafel plots for the electrochemistry, that is, how the kinetic overpotentials for discharge ηdis and charge ηchg vary with the current i (or current density j). The separation between the Tafel plots for discharge and charge defines the maximum possible electrical efficiency for a discharge−charge cycle at a given i. It is well known that Galvansostatic discharge curves of typical Swagelok or coin-type Li−O2 batteries show a currentdependent voltage plateau Udis, followed by a rapid decrease in © 2013 American Chemical Society

Received: January 4, 2013 Accepted: January 30, 2013 Published: January 30, 2013 556

dx.doi.org/10.1021/jz400019y | J. Phys. Chem. Lett. 2013, 4, 556−560

The Journal of Physical Chemistry Letters

Letter

Figure 1. Measurement of Li−O2 discharge potential in the plateau region (Udiss) as a function of Galvanostatic current i imposed during discharge. Swagelok cells are used with Li metal anode, 1 M LiN(CF3SO2)2 in dimethoxyethane (LiTFSI/DME) electrolytesoaked Celgard separators (∼50 μm thick) and either P50 carbon paper cathode (red triangles) or XC72 C with PTFE binder on SS mesh as cathode (blue squares). The superficial area of the cathode is ∼1 cm2 with a total cathode active surface area of ∼0.1 m2.

circuit prior to discharge. Although there are many contributions to the cell impedance, we believe that the dominant one is due to the solid electrolyte interface at the Li metal anode. The results in Figure 1 are typical for Swagelok or coin-type cells and therefore fully mask the fundamental electrochemistry. To measure the kinetic overpotentials that provide ultimate limits to discharge−charge cycle efficiency, the cell iR drop must be eliminated. To achieve this condition, we measured Galvanostatic discharge and charge experiments in rapidly stirred three-electrode bulk electrolysis cells, with a flat polished glassy carbon (GC) electrode of surface area ∼1 cm2 as the working electrode and Li metal as the counter and reference electrodes. (See the SI for details of the experiments.) The low current into the reference electrode minimizes the iR drops (with additional iR compensation to remove any residual secondary effects), and the rapid stirring minimizes Li+ and O2 transport limitations (concentration polarizations). Therefore, the fundamental kinetic overpotentials can be measured in the bulk electrolysis cell over a very wide range of current densities. Figure 2a shows Galvanostatic discharge plots (output cell potential U vs discharge capacity Qdis) of Li−O2 at various current densities j in the bulk electrolysis cell using the same electrolyte as in Figure 1. (See the SI.) Previous experiments have shown that Li2O2 forms a continuous film on the GC surface as a result of discharge.7 At a given j, the discharge plot shows an initial drop in potential due to the kinetic overpotential ηdis, a linear decrease in potential with Qdis ascribed to an jR drop through the thickening Li2O2 film, and a “sudden death” characterized by a rapid decrease in U. The latter two phenomena have previously been described as charge-transport limitations through the Li2O2 film.7 In this note, we discuss only the initial drop in U at Qdis ≈ 0 and its j dependence. We define Udis(Qdis = 0) as the linear extrapolation of the linear part of the discharge curve (jR drop region) to Qdis = 0. This defines the kinetic overpotential ηdis (when U0 is known) without distortions due to charge transport limitations in the regime for Li2O2 growth on Li2O2 that we believe dominates the electrochemistry during battery discharge.7,8

Figure 2. (a) Output potential during Li−O2 galvanostatic discharge in the bulk electrolysis cell at the varying current densities given in the legend. Li metal is used as both the reference and counter electrodes and polished glassy carbon (GC) is the working electrode. All electrodes have ∼1 cm2 surface area. 1 M LiTFSI/DME is the electrolyte. (b) Output potential during Galvanostatic charging in the bulk electrolysis cell at the varying current densities given in the legend following a 0.3 μAh Li−O2 discharge at 20 μA/cm2 current density as in panel a.

Figure 2b shows galvanostatic charging (U vs charging capacity Qchg) for various charging currents, all following a Galvanostatic discharge of Qdis = 0.33 μAh at a current density of 20 μA/cm2. This Qdis is ∼1/3 of that for sudden death at 20 μA/cm2. To maintain the constant charging current, U increases with Qchg. We have previously presented evidence that this increase in potential is not due to a fundamental kinetic overpotential but simply results from the build up of contamination products at the electrolyte−Li2O2 interface.4 We define the initial U defined by a linear extrapolation to Qchg = 0 from the first ∼0.05 μAh charging as a measure of initial charging potential for a given discharge capacity, Uchg(Qdis). Because there is a jR drop in the Li2O2 film during charging just as in discharge, Uchg(Qdis) depends on the thickness of the Li2O2 film initially deposited or Qdis. To correct for this, similar charging measurements were also made following discharges of Qdis = 0.16 and 0.50 μAh and the values for Uchg(Qdis) at a given current linearly extrapolated to Qdis = 0. We consider these extrapolations Uchg(Qdis = 0) as the fundamental measurement of the charging kinetic overpotential ηchg (when U0 is known). 557

dx.doi.org/10.1021/jz400019y | J. Phys. Chem. Lett. 2013, 4, 556−560

The Journal of Physical Chemistry Letters

Letter

⎛ ΔG0± ⎞ ⎛ αe|Udis − U0| ⎞ j ∝ −[Li+]O2* exp⎜ − ⎟ exp⎜ ⎟ kBT ⎝ kBT ⎠ ⎝ ⎠

Figure 3a shows the Tafel plot for discharge and charge, that is, log(j) versus Udis(Qdis = 0) and Uchg(Qdis = 0) as defined

(1)

where [Li ] and O2* are the reactant at or near the surface, ΔG0± is the kinetic barrier to the limiting reaction at the equilibrium potential, α is the symmetry factor, and e is the charge on the electron. ΔG0± is commonly referred to as the desolvation barrier. This equation implies that for any fixed U, j depends linearly on O2* and hence on the headspace pressure of O2, P. O2* follows a Langmuirian dependence on P at a single fixed U,9 and Figure S2(a) in the Supporting Information shows that this is true for all U. For Galvanostatic discharge at fixed j, the overpotentials for discharge |Udis − U0| will also depend on T, and this variation measures ΔG0±/α. Tafel plots for discharge at two different currents, 30 and 50 C, are given in Figure S2(b) in the Supporting Information. Analysis of this plot in terms of the Tafel equation (and an equivalent set of measurements for 20 and 40 C) give a value ΔG0±/α = 1.5 ± 0.2 with U0 = 2.85 V. With α ≈ 0.5, this yields a barrier of ∼0.75 eV. As previously pointed out, during discharge a crystalline deposit of Li2O2 is formed. Therefore, a theoretical treatment of the current−voltage response of the Li−O2 battery requires a detailed understanding of crystal growth and dissolution of Li2O2 for discharge and charge, respectively. Whereas the nucleation/dissolution on the cathode material itself is important at the outset of discharge/end of charge, the growth and dissolution of Li2O2 on the Li2O2 itself dominates the overall discharge/charge processes. In previous work, we have emphasized that growth/dissolution proceeds through sequential transfers of (Li+ + e−), with potential limiting steps of (Li+ + e−) + O2* → LiO2* for crystal growth and Li2O2 → LiO2* + (Li+ + e−) for dissolution.7−9 These charge transfers can occur on different crystal facets, on different terminations on those facets, and at different sites (terrace, step or kink) and could involve different combinations of nucleation and diffusion. Thus, the overall mechanism for growth/dissolution can be quite complicated. The surface energies of the most stable facets and terminations are given in Figure S3 in the Supporting Information. While the surface energies of the stoichiometric surfaces are potential-independent, the surface energy of nonstoichiometric O-rich surfaces is U-dependent. A thermodynamic analysis of growth and dissolution on some of the possible paths (different stable surface structures and terminations and the different charge-transfer sites) has been presented elsewhere.8 Using density functional theory calculations, the theoretical limiting potentials for discharge and charge for a given path are defined as the highest (lowest) potential at which all of the discharge (charge) steps are downhill in free energy. This work focused principally on some (but not all) paths that have a low overpotential at a given facet/termination/charge transfer site because these will dominate the activity due to the exponential dependence of the rate on the overpotential. The overpotentials for the paths calculated in ref 8 are given in Table 1. This thermodynamic treatment predicts a lower limit to the overpotentials associated with the discharge and charge for a given path.10 Although desolvation barriers could, in principle, cause higher kinetic overpotentials, the modest desolvation barrier of ∼0.75 eV should increase these minimally for the current densities described here. The validity of this thermodynamic treatment in predicting overpotentials in agreement with experiments is well-documented for the +

Figure 3. (a) Experimental Tafel plots for Li−O2 discharge (ORR, blue triangles) and for charging following discharge (OER, red squares). (b) First-principles theoretical prediction of the Tafel plots for Li2O2 crystal growth, the dominant process in discharge (blue triangles), and Li2O2 dissolution, the dominant process in charge (red squares).

above. Once the surface is covered with Li2O2, the open circuit or equilibrium potential U0 ≈ 2.85 V (vs Li/Li+). This Figure shows that the kinetic overpotentials, for both discharge and charge, are very small. The separation between Udis(Qdis = 0) and Uchg(Qdis = 0) at 10 μA/cm2, a current density equivalent to a 10 mA Li-air battery with a cathode of ∼0.1 m2 active surface area in 1 cm2 superficial area, is