High-rate electrochemical energy storage through Li+ intercalation pseudocapacitance

LETTERS PUBLISHED ONLINE: 14 APRIL 2013 | DOI: 10.1038/NMAT3601 High-rate electrochemical energy storage through Li+ intercalation pseudocapacitance Veronica Augustyn1, Jérémy Come2,3, Michael A. Lowe4, Jong Woung Kim1, Pierre-Louis Taberna2,3, Sarah H. Tolbert5, Héctor D. Abruña4, Patrice Simon2,3 and Bruce Dunn1* Pseudocapacitance is commonly associated with surface or near-surface reversible redox reactions, as observed with RuO2 · xH2O in an acidic electrolyte. However, we recently demonstrated that a pseudocapacitive mechanism occurs when lithium ions are inserted into mesoporous and nanocrystal films of orthorhombic Nb2O5 (T-Nb2O5; refs 1,2). Here, we quantify the kinetics of charge storage in T-Nb2O5: currents that vary inversely with time, charge-storage capacity that is mostly independent of rate, and redox peaks that exhibit small voltage offsets even at high rates. We also define the structural characteristics necessary for this process, termed intercalation pseudocapacitance, which are a crystalline network that offers two-dimensional transport pathways and little structural change on intercalation. The principal benefit realized from intercalation pseudocapacitance is that high levels of charge storage are achieved within short periods of time because there are no limitations from solid-state diffusion. Thick electrodes (up to 40µm thick) prepared withT-Nb2O5 offer the promise of exploiting intercalation pseudocapacitance to obtain high-rate charge-storage devices. Pseudocapacitance occurs whenever the charge (Q) depends on the change in potential (dE), yielding a capacitance (dQ/dE) (ref. 3). The capacity can be due to monolayer adsorption of ions at an electrode surface, as in the underpotential deposition of metals4; surface redox reactions as in RuO2; or ion intercalation that does not result in a phase change. Although these redox processes are Faradaic in nature, their phenomenological behaviour, and response to experimental variables such as sweep rate, are those typical of capacitors. All of these scenarios produce a relationship between the fractional extent of charge storage, X , and the potential of the form3,5: E ∼ ( RT nF ) ln [ X (1−X) ] (1) whereR is the ideal gas constant (Jmol−1 K−1),T is the temperature (K), F is Faraday’s constant (96,485 Asmol−1) and n is the number of electrons involved in the reaction. In all of these cases, a constant-current experiment yields a potential E that changes with the extent of chargeQ according to: Q=C1E where Q is the charge passed (Coulombs), 1E is the potential change (V) and C is the pseudocapacitance (F). This behaviour 1Department of Materials Science and Engineering, University of California, Los Angeles, California 90095, USA, 2Department of Materials Science, Université Paul Sabatier, CIRIMAT UMR CNRS 5085, Toulouse 31062, France, 3Réseau sur le Stockage Electrochimique de l’Energie (RS2E), FR CNRS 3459, France, 4Department of Chemistry and Chemical Biology, Cornell University, Ithaca, New York 14853, USA, 5Department of Chemistry and Biochemistry, University of California, Los Angeles, California 90095, USA. *e-mail: bdunn@ucla.edu. is typical of capacitive charge/discharge, thus leading to the term pseudocapacitance3. Of the three pseudocapacitive mechanisms mentioned above, underpotential deposition and surface redox reaction pseudoca- pacitance exhibit kinetics indicative of surface-controlled electro- chemical processes6: i=Cv where i is the current (A) and v is the sweep rate (mV s−1) of a cyclic voltammetry experiment. However, in intercalation pseudocapacitance, as described herein, charge storage does not occur on the surface but in the bulk material. The kinetics are not diffusion-limited and instead are limited by surface processes so that the overall behaviour seems capacitive. Intercalation pseudocapacitance is rarely observed because in most intercalation materials charge storage (even in thin films, as in anatase TiO2; refs 7,8) is limited by solid-state diffusion and therefore the peak currents scale with v1/2. Here we investigate the phenomenon of intercalation pseudo- capacitance and the high-rate behaviour of T -Nb2O5 using two different electrode techniques that provide a wide variation in sweep rates. For timescales between ∼3 h and 60 s (sweep rates of 0.1–20mV s−1 within a voltage window of 1.2 V), we used a thin-film electrode. For shorter timescales where ohmic polariza- tion is significant (60–500mV s−1), we used a cavity microelectrode where the active material was mixed with a conductive carbon black to alleviate the loss of electrical transport (ohmic losses)9. To confirm the small ohmic drop of this electrode, we performed cyclic voltammetry from 100–500mV s−1 in a bulky-ion electrolyte (Sup- plementary Fig. S1). In addition, we prepared thick films (∼40 µm) of T -Nb2O5 to determine whether the high-rate capability was limited to thin films. Charge storage from the intercalation of lithium ions intoNb2O5 can be expressed as: Nb2O5+xLi++xe−↔ LixNb2O5 where the maximum capacity is x = 2 (ref. 10). Figure 1a shows cyclic voltammograms from 100 to 500mV s−1 in a cavity microelectrode where it is evident that both anodic and cathodic peaks are broad, about 600mV. There is also a noticeable peak shift (and increase in peak separation 1Ep) as the sweep rate increases, but the capacity remains reversible. Figure 1b presents a plot of log(i) versus log(ν) from 0.1 to 500mV s−1 for both cathodic and anodic peaks. Assuming that the current obeys a power-law NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials 1 © 2013 Macmillan Publishers Limited. All rights reserved. Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 LETTERS NATURE MATERIALS DOI: 10.1038/NMAT3601 ¬1 0 1 2 31.0 1.5 2.0 2.5 3.0 0.0 1 × 10¬3 0.0 0.2 0.4 0.6 0.8 1.0 50¬500 mV s¬1 Cathodic: 0.8 Anodic: 0.7 a b Cathodic Anodic lo g( pe ak c ur re nt , A ) log(sweep rate, mV s¬1) 0.1¬20 mV s¬1 Cathodic: 1 Anodic: 1 C ur re nt ( m A ) Potential (V versus Li/Li+) N or m al iz ed c ap ac ity Sweep rate¬1/2 (s1/2 mV¬1/2) Region 1 <20 mV s¬1 Region 1 <20 mV s¬1 C at ho di c pe ak s hi ft ( V ) ¬1 × 10¬3 100 mV s¬1 200 mV s¬1 300 mV s¬1 400 mV s¬1 500 mV s¬1 ¬6 ¬5 ¬4 ¬3 ¬2 ¬1 0 1 2 3 log(sweep rate, mV s¬1) 0.0 0.2 0.4 0.6 0.8 1.0 Re gio n 2 >2 0 m V s¬ 1 Re gio n 2 >2 0 m V s¬ 1 0.0 0.1 0.2 0.3 0.4 0.5c d Figure 1 | Kinetic analysis of the electrochemical behaviour of T-Nb2O5. a, Cyclic voltammograms from 100 to 500 mV s−1 demonstrate the high-rate capability of the material. b, b-value determination of the peak anodic and cathodic currents shows that this value is approximately 1 up to 50 mV s−1. This indicates that even at the peak currents, charge storage is capacitive. c, Capacity versus v−1/2 allows for the separation of diffusion-controlled capacity from capacitive-controlled capacity; two distinct kinetic regions emerge when the sweep rate is varied from 1 to 500 mV s−1. The dashed diagonal line corresponds to the extrapolation of the infinite sweep rate capacitance using the capacity between 2 and 20 mV s−1. d, The variation of the cathodic peak voltage with the sweep rate exhibits a region of small peak separation followed by increased separation at 20 mV s−1, and represents another method of identifying systems with facile intercalation kinetics. relationship with the sweep rate leads to8: i= avb where a and b are adjustable values. Whereas a b-value of 0.5 would indicate that the current is controlled by semi-infinite linear diffu- sion, a value of 1 indicates that the current is surface-controlled. For sweep rates ranging from 0.1 to 20mV s−1, corresponding to charging times >60 s, the b-value for both the cathodic and anodic peaks is 1, indicating that the kinetics are surface-controlled, and thus fast. Figure 1b also exhibits a change in the slope of the anodic and cathodic peak currents at∼50mV s−1. This change in slope cor- responds to a decrease in b-value to 0.8 and 0.7 for the cathodic and anodic currents, respectively, at sweep rates >50mV s−1 (charging times <20 s). This limitation to the rate capability can arise from numerous sources including an increase of the ohmic contribution (active material resistance, solid–electrolyte interphase resistance) and/or diffusion constraints/limitations11. In the limit of slow diffusion, bwould approach a value of 0.5 as described above. The relationship between capacity and sweep rate can also establish the rate-limiting step of a charge-storage mechanism12. In a plot of Q versus v−1/2, regions that are linear represent capacity limited by semi-infinite linear diffusion whereas capacitive contributions are independent of the sweep rate. At sweep rates below 20mV s−1, the extrapolated y-intercept yields the infinite sweep rate capacitance13. Figure 1c shows the plot of normalized capacity versus v−1/2 for T -Nb2O5 from 1 to 500mV s−1 (the gravimetric capacity for the thin-film electrode is shown in Supplementary Fig. S2). Analogous to the behaviour of the peak current in Fig. 1b, there are two distinct regions in Fig. 1c. In region 1, at sweep rates <20mV s−1, the capacity is mostly independent of sweep rate. The magnitude of the capacity is ∼130mAh g−1 or ∼65% of the theoretical value based on a two-electron redox reaction with Nb2O5. In this range, solid- state lithium-ion diffusion is not the rate-limiting step for charge storage. In region 2, from 50 to 500mV s−1, the capacity decreases linearly with v−1/2. This indicates that charge storage is mainly diffusion-controlled at high sweep rates. That is, for charging times 2 NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials © 2013 Macmillan Publishers Limited. All rights reserved. Administrator 高亮 Administrator 高亮 Administrator 矩形 Administrator 备注 注意这个讨论关于电流和扫描速率的关系 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 NATURE MATERIALS DOI: 10.1038/NMAT3601 LETTERS 0.0 0.5 1.0 1.5 a b Li4Ti5O12 Nb2O5 C ap ac ity ( m A h g¬ 1 ) C rate Po te nt ia l ( V v er su s Li /L i+ ) Capacity (mAh g¬1) 0 50 100 150 x in LixNb2O5 1.2 1.5 1.8 2.1 2.4 2.7 3.0 1 10 100 1,000 0 20 40 60 80 100 120 140 160 180 200 Insertion Deinsertion Figure 2 | Electrochemical cycling of a 40-µm-thick T-Nb2O5 electrode. a, Galvanostatic cycling of a thick Nb2O5 electrode at a 10C rate. b, Comparison of the rate capability of T-Nb2O5 with a high-rate lithium-ion anode, Li4Ti5O12, at various C-rates (Li4Ti5O12 data reproduced from ref. 16). of <20 s, diffusion is rate-limiting, similar to most traditional battery electrodes. However, for charging times of 1min (60C) or longer, there is no indication of diffusion limitations and this intercalation-based systembehaves in a fully capacitivemanner. Another feature of T -Nb2O5 at sweep rates <20mV s−1 is that the peak voltage shifts with sweep rate are small (Fig. 1d). The cathodic peak shift is <0.1V at sweep rates below 10mV s−1. As a result, the anodic and cathodic peaks overlap at 0.1mV s−1 (Supplementary Fig. S3) and it is in this behaviour that the similarity to surface redox reactions is most apparent14. In many lithium-ion intercalation materials, the peak separation is significant even in thin films and at slow sweep rates (for example, 1Ep = 0.13V for LiCoO2 at 0.1mV s−1; ref. 15). This type of behaviour is often associated with crystallographic phase changes during the Faradaic process, and contrasts with intercalation materials that form a solid solution, such as T -Nb2O5. Besides identifying facile intercalation, the peak voltage separation is related to the high-power capability of a material. As the charging time decreases, that is, at higher current densities, the peak separation in a battery material increases owing to polarization (reflecting the higher overpotentials necessary to deliver the higher currents), so that at higher rates the energy required to fully charge the material is significantly larger than the energy available on discharge. The high-rate behaviour of T -Nb2O5 is not limited to thin films or to experiments with small amounts of active material. The constant-current charge/discharge of a 40-µm-thick T -Nb2O5 (1mg cm−2) electrode at a 10C rate is shown in Fig. 2a. At this rate, the capacity is 130mAh g−1 and E varies linearly with Q as expected for a pseudocapacitive process from equation (1). This represents capacities typical of battery materials but at rates closer to those of supercapacitors. The rate capability of T -Nb2O5 from 1 to 1,000C is shown in Fig. 2b and compared with that of Li4Ti5O12 (of comparable electrode dimensions)16. Li4Ti5O12 is chosen as an example of a high-rate lithium-ion anode material. The rate capability for T -Nb2O5 is significantly better than Li4Ti5O12 above 30C and even at a 1,000C rate the capacity of the thick T -Nb2O5 electrode is∼40mAh g−1. The thick electrode results verify that the intercalation pseudocapacitance mechanism is not due to thin-film or surface effects, such as vacancies or contributions of the first few atoms from the surface. Indeed, even for the thick electrode, the b-values for the anodic and cathodic peak currents are 1 from 1 to 10mV s−1 (Supplementary Fig. S4). The high-rate capability of T -Nb2O5 implies that the crystal structure permits exceptionally rapid ionic transport. As shown in Fig. 3a, the unit cell has sheets of edge- or corner-sharing distorted polyhedra lying parallel to the (001) plane, with eachNb5+ surrounded by either 6 or 7 O2−. The polyhedra are exclusively corner-sharing along the [001] direction with 5% of the Nb5+ ions randomly located in 9-coordinate sites between the (001) poly- hedral planes17. The mostly empty octahedral sites between (001) planes provide natural tunnels for lithium-ion transport through- out the a–b plane. Calculations indicate that the (001) plane allows degenerate pathwayswith low energy barriers for ion transport18. A previous in situ X-ray diffraction study showed that the insertion of lithium into T -Nb2O5 results in a solid solution with no apparent phase changes19, and negligible changes to lattice constants2 and unit-cell volume20 up to ∼1.25 Li+/Nb2O5. In situ X-ray absorption spectroscopy (XAS) demonstrated that lithiation reduces Nb5+ to Nb4+ (ref. 19). The in situ XAS studies carried out here confirm that lithiation results in a continuous change in oxidation state (Fig. 3b) and the Fourier-transform of the extended X-ray absorption fine structure (EXAFS) indicates that the insertion reaction proceeds through two stages (Fig. 3c). From 2.5 to 1.75 V, the EXAFS signal from the various Nb–O bond lengths in Nb2O5 (1.40–1.85 Å) merges to a single peak at an intermediate bond length (1.75 Å), indicating that lithiation increases the Nb-centred symmetry. The lithiation is probably faster at low Li+ levels20 owing to greater availability of sites and less interaction between cations. At lower potentials, the new EXAFS peak shifts to longer bond distances (1.85 Å) as a consequence of increased Li–O interactions at higher Li+ content. These structural studies emphasize the value of an open, layered structure to enable rapid ion transport within the active material. The results presented in this study establish that T -Nb2O5 exhibits electrochemical features of a pseudocapacitive material despite charge storage occurring in the bulk. Such behaviour is consistent with intercalation pseudocapacitance. The electrochem- ical features indicative of this mechanism are currents that are linearly proportional to the sweep rate, capacity that does not vary significantly with charging time, and peak potentials that do not shift significantly with sweep rate. A key design rule for intercalation pseudocapacitance at the atomic scale is a structure that does not undergo phase transformations on intercalation. In addition, facile two-dimensional (2D) lithium-ion diffusion pathways are NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials 3 © 2013 Macmillan Publishers Limited. All rights reserved. Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 备注 高扫描速率下，电位增宽，属于极化 Administrator 高亮 LETTERS NATURE MATERIALS DOI: 10.1038/NMAT3601 18,980 19,000 19,020 19,040 1 2 3 4 D er iv at iv e, n or m al iz ed a bs or pt io n Incident energy (eV) a a k2 | (R )| ( A ng ¬ 3 ) χ R (Å) b c c 2.50 V 1.80 V 1.65 V 1.50 V 1.35 V 0.0 0.5 1.0 1.5 Figure 3 | Structural features of lithium intercalation in T-Nb2O5. a, The structure of T-Nb2O5 stacked along the c axis demonstrates the layered arrangement of oxygen (red) and niobium (inside polyhedra) atoms along the a–b plane. b, Derivative of Nb K-edge X-ray absorption near-edge spectra at selected cell voltages, showing a systematic shift to lower energies as Nb5+ is reduced to Nb4+. c, k2-weighted Fourier-transformed Nb K-edge EXAFS at selected cell voltages. Table 1 |Comparison between charge storage in two different pseudocapacitive materials. RuO2 ·xH2O T-Nb2O5 Electrolyte H+, aqueous21 Li+, non-aqueous Structure Amorphous/ nanocrystalline22,23 Crystalline Structural water Necessary for proton conduction22 Not necessary Charge storage mechanism Surface/near- surface23,24 Bulk Pseudocapacitance type Surface redox Intercalation RuO2 ·xH2O and T-Nb2O5 both exhibit capacitive behaviour. The mechanism of charge storage, however, is different and results in different structural requirements for high capacitance. important. Charge storage that behaves as a quasi-2D process ex- hibits similar behaviour to 2D surface adsorption reactions14. These features contrast with those of pseudocapacitive RuO2 ·xH2Owhere charge storage occurs mainly on the surface or near-surface21–24 as summarized in Table 1. The results here are exciting because they demonstrate that for charging times as fast as 1min (60C rate), there are no diffusion limitations in T -Nb2O5. As the high-rate capability is due to fast ion diffusion in the bulk, this mechanism may be very good for thick electrodes because surface exposure to the electrolyte is not critical. To achieve devices with high energy density, further engineering at the nanoscale and beyond will be necessary to preserve the atomic-scale behaviour observed in thin films and microelectrodes. In particular, maintaining proper electronic conduction pathways will be critical. Methods Synthesis. The synthesis of T -Nb2O5 nanocrystals was reported previously2. Briefly, 2.56mmol of NbCl5 (Sigma-Aldrich) was dissolved in 2ml of ethanol (Fisher Scientific). In a separate vial, 0.23ml of deionized water was mixed with 2ml of ethanol. Both vials were then chilled for 2 h. The two solutions were then mixed together while 1ml of propylene oxide (Sigma-Aldrich) was slowly added, forming a transparent gel. This gel was aged for 1 day and then soaked in acetone (Fisher Scientific) for 5 days. After supercritical drying with CO2, the gel was transformed into an amorphous Nb2O5 aerogel. Crystallization to the T -phase occurred by heat treatment at 600 ◦C for 2 h in air. Characterization. Thin-film electrodes were fabricated by drop-casting a well-sonicated solution of T -Nb2O5 in ethanol onto an oxygen plasma etched stainless-steel foil (Alfa Aesar). For thin-film measurements, lithium foils (Sigma-Aldrich) served as the reference and counter electrodes, and the electrolyte was 1M LiClO4 (Sigma-Aldrich) in propylene carbonate (Sigma-Aldrich). The microelectrode preparation w