Electrochemical characterization of lithium cobalt

Electrochimica Acta 281 (2018) 822e830

Contents lists available at ScienceDirect

Electrochimica Acta journal homepage: www.elsevier.com/locate/electacta

Electrochemical characterization of lithium cobalt oxide within aqueous flow suspensions as an indicator of rate capability in lithiumion battery electrodes Linxiao Geng, Matthew E. Denecke, Sonia B. Foley, Hongxu Dong, Zhaoxiang Qi, Gary M. Koenig Jr. * Department of Chemical Engineering, University of Virginia, 102 Engineers Way, Charlottesville, VA 22904-4741, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 23 March 2018 Received in revised form 14 May 2018 Accepted 5 June 2018 Available online 6 June 2018

Rate capability is a critical metric for assessing lithium-ion battery materials, especially in the electric vehicle industry which requires high battery power density. The typical method for rate capability evaluation of battery materials involves electrode processing, battery assembly, and testing on a battery cycler with varying current rates. This method is very important in evaluating battery materials, however, it is time-consuming and depending on the cell type can take days to weeks. Rate capability can also be affected by other factors besides the active material properties including electrode slurry homogeneity, contact junctions formed during battery assembly, and the connectivity of the interfaces within the battery electrodes. Herein, we will describe the use of a technique termed dispersed particle resistance (DPR) to evaluate the battery cathode active material LiCoO2 (LCO), the first layered oxide evaluated using this method. This new technique works relatively fast, providing qualitative analysis within several minutes, and does not require electrode fabrication or additional materials in the system besides a carrier fluid for the electrolyte. While DPR does not explicitly give rate capability of the materials, DPR was successfully demonstrated to provide the relative rate capability for multiple different LCO samples, as supported by the results obtained from conventional rate capability testing using cycling data. Additionally, the physical properties of the material probed by the DPR were investigated by determining the lithium diffusivity in the crystal structure of the different LCO materials with galvanostatic intermittent titration technique. © 2018 Elsevier Ltd. All rights reserved.

Keywords: Dispersed particle resistance Flowing particle analysis Battery material characterization Lithium-ion battery Particle suspension

1. Introduction The demand for long lasting and high-power-density lithiumion (Li-ion) batteries continues to increase concurrent with the increased demand for electric vehicles and portable electronics [1,2]. Battery researchers and Li-ion battery material manufacturers have explored and developed a wide variety of active material structures and compositions in an effort to improve battery energy and power density metrics, including lithium iron phosphate (LiFePO4, LFP), lithium cobalt oxide (LiCoO2, LCO), and lithium nickel manganese cobalt oxide (LiNixMn1-xCo1-x-yO2, NMC) [3e11].

* Corresponding author. E-mail addresses: [email protected] (L. Geng), [email protected] (M.E. Denecke), [email protected] (S.B. Foley), [email protected] (H. Dong), [email protected] (Z. Qi), [email protected] (G.M. Koenig). https://doi.org/10.1016/j.electacta.2018.06.037 0013-4686/© 2018 Elsevier Ltd. All rights reserved.

Determining active material electrochemical properties is essential both for materials researchers to understand the structureproperty relationships of battery materials and for battery manufacturers to validate active material quality before cell assembly [12,13]. Rate capability is one of the most important metrics of cathode materials because it is directly related to the power capability of the resulting battery cells and packs [14,15]. Typically, the rate capability of a cathode material is evaluated by processing the active material into a composite electrode and then using the electrode during fabrication of a battery cell for electrochemical testing. The overall process includes mixing electrode materials with conductive carbon, an appropriate binder, and solvent, casting slurry onto a current collector to form an electrode, assembling a battery with the electrode along with all other necessary components, and finally cycling the battery using different rates of charge/ discharge. This procedure is very time-consuming, usually taking several days. In addition, the results can be ambiguous because they

L. Geng et al. / Electrochimica Acta 281 (2018) 822e830

are greatly affected by factors including slurry homogeneity, binder integrity, connectivity of cell components, conductive carbon loading percentage, and other factors in addition to the intrinsic properties of the active material [16e19]. Consequently, a new technique that directly provides, or at least correlates to, rate capability in a more time efficient process would be desirable to reduce the time required to compare the electrochemical performance of active materials or to validate that a given material meets a specified rate capability metric. Few methods have been proposed to evaluate the rate capability of a cathode material beyond the aforementioned conventional method of charge/discharge cycling at varying rates. Gabersceck et al. proposed implementing mass electrode resistance, or RU, to evaluate the rate capability of cathode materials of the same chemistry [20]. They demonstrated a linear relationship between the overpotential and the current density (per mass) by observing the potential of coin cells at the same state of charge at increasing current densities. The slope of the regression line is defined as mass electrode resistance (RU). The authors successfully correlated the RU of a material with its rate capability since the rate capability of a material is closely related to how the overpotential responds with different current densities. Despite the functionality of this method, assembly and cycling of batteries were still required in order to determine RU. Microelectrode techniques have also been used to determine the rate capability and kinetic properties of Li-ion battery materials based on single particle analysis. This technique did not require electrode fabrication with carbon additives and polymer binders and was thus active material specific. However, it required a complex experimental setup and does not sample a large ensemble of active material in a high throughput manner [21e23]. In previous reports, our group designed an electrochemical system where battery material particle suspensions in electrolytes were either agitated in a beaker or flowed through a channel containing a current collector [24,25]. This technique was based on a concept of electrochemical reactions that occur only when solid particles in a dispersion were in contact with an electrode or current collector [26e28]. By applying discrete sequential chronoamperometry, a single parameter was extracted termed dispersed particle resistance (DPR). As a proof of concept, DPR was successfully correlated with rate capability using lithium titanium oxide (LTO) suspensions in organic Li-ion battery electrolyte [24]. The DPR method does not require the assembly and cycling of a battery which meant it was relatively fast, being completed in the first report in less than twenty minutes. The preparation was also very simple, involving mixing active material powder with electrolyte and stirring the resulting suspension in a beaker. This new technique may have promise in reducing the cost of labor, time, and equipment for rate capability comparisons between materials. In a follow-up report, a new cell design was incorporated that a flow-through geometry, and a number of different LFP cathode powders were used for comparative analysis. Furthermore, the organic electrolyte was replaced with a Li2SO4 aqueous electrolyte, which enabled the cell apparatus to be moved out of the glovebox, further enhancing the flexibility of the method [25]. The measured DPR in the flowthrough system also provided a metric that correlated with the relative rate capability of the LFP materials as confirmed by conventional battery cycling and rate capability analysis. In this paper, the DPR technique is further extended to cathode materials with a layered structure, which are the most common materials used in commercial Li-ion battery cathodes. The material that was used as an exemplary layered oxide material was LCO. The first application of DPR using an aqueous carrier fluid was with LFP cathode material, and LFP was a desirable initial cathode material because of its relatively low redox potential compared to other cathode materials and its previously demonstrated suitability as an

823

aqueous Li-ion cathode material [24,27e30]. In contrast, layered oxide materials have higher redox potentials and previously have been reported to have stability limitations when in contact with water [31,32], and thus it was not clear if DPR with an aqueous carrier fluid would be suitable for these materials. This paper will demonstrate that aqueous DPR measurements correlate to the relative rate capability of a series of LCO cathode materials. These LCO materials were synthesized using different processing conditions and characterized with a variety of different techniques, with galvanostatic intermittent titration technique (GITT) in particular providing evidence that the DPR, and rate capability, were most sensitive to the diffusivity of the Liþ in the synthesized materials. 2. Experimental 2.1. Materials synthesis and characterization LCO samples were synthesized with a method previously reported [5]. Cobalt oxalate dihydrate (CoC2O42H2O) was synthesized via coprecipitation: equal volume of 0.1 M ammonium oxalate monohydrate ((NH4)2C2O4H2O, Sigma Aldrich 99.5e101.0%) and 0.1 M cobalt nitrate hexahydrate (Co(NO3)26H2O, Fisher Scientific reagent grade) solutions were mixed under agitation at 60 C to yield CoC2O42H2O. The pink solid particle precipitates were collected by vacuum filtration and dried thoroughly at 80 C. Then CoC2O42H2O was converted to LCO via calcination in a mixture with lithium hydroxide (LiOH, Fisher Scientific lab grade), where the LiOH was added to achieve 5% excess lithium (Li:Co ratio of 1.05:1) in a Carbolite CWF 1300 box furnace. The four target calcination temperatures selected were 750, 800, 850 and 900 C. The calcination procedure was to heat to the target temperature at a ramp rate of 1 C min1, and then upon reaching the target temperature the furnace was shut off and allowed to cool to ambient without control over the cooling rate. The LCO products were then milled on a roller (US Stoneware) at 100 rpm via a wet soft milling procedure with zirconia beads and 1-Butanol solvent, yielding four samples used in this paper with the names LCO750, LCO800, LCO850 and LCO900, respectively (with each number reflecting the target calcination temperature). The morphology of the synthesized samples was determined with a Quanta 650 scanning electron microscope (SEM). A Panalytical X'pert diffractometer with a Cu Ka radiation source was used to obtain the X-ray diffraction (XRD) patterns of the LCO samples. BET surface area of the LCO samples were determined with a NOVA 2200e. 2.2. Active material suspension electrochemical evaluation The DPR experiments were performed in a custom flow cell, a schematic of which can be seen in the Supporting Information, Fig. S1. The working electrode was a serpentine gold wire (0.25 mm diameter and 60 cm long). The counter electrode was a platinum wire (5 mm diameter and 8 cm long). An Ag/AgCl electrode (Pine instruments) served as the reference electrode. The suspension that was brought into contact with the working electrode via flowing through a channel was made by mixing 2 vol% LCO powders with 1 M Li2SO4 aqueous electrolyte. The counter electrode was in contact with 1 M Li2SO4 within a channel and did not contain any other material or particles. The anode and cathode channels were separated by a polymer separator (Celgard® 3401). During DPR analysis, the LCO suspensions were agitated by a magnetic stir bar at 600 rpm and pumped through the cathode channel at a rate of 80 ml min1 using a MasterFlex peristaltic pump (Cole-Parmer). All the electrochemical experiments on this custom device were performed with a Biologic SP-150 potentiostat.

824


2.3. Coin cell fabrication and electrochemical characterization CR-2032 type coin cells were used to evaluate the rate capability of the LCO materials within composite electrodes, as well as to conduct GITT analysis. In these coin cells, LCO electrodes served as the cathode (working electrode) with lithium foil as the anode (counter and reference electrode). The separator used was Celgard® 2325. The cathode was fabricated by coating LCO slurry onto aluminum foil current collector using a doctor blade with a gap thickness of 200 mm. The slurry was made by mixing 80 wt% LCO, 10 wt% carbon black, and 10 wt% of polyvinylidene fluoride in NMethy-2-pyrrolidone solvent via a mechanical mixer for 10 min. The electrode was dried in an oven at 80 C overnight and further dried in a vacuum oven at 80 C for an additional three hours. Electrode disks with a diameter of 9/16 inch were punched, and the typical loading of LCO on a single electrode was approximately 10 mg. The electrolyte used in the coin cells was 1.2 M lithium hexafluorophosphate in ethylene carbonate:ethyl methyl carbonate with a volume ratio of 3:7 (BASF Corporation). The coin cells were assembled in an argon-filled glovebox with concentrations of both O2 and H2O less than 1 ppm. The galvanostatic charge/ discharge and GITT were performed on a MACCOR battery cycler. For the GITT analysis, three conditioning cycles were carried out at a rate of 0.2 C over a voltage range of 2.5 Ve4.2 V. On the fourth cycle, the GITT program was set in the following way: each current pulse lasted for 10 min at a rate of 0.2 C, and after each current pulse the cells were allowed to rest for 1.5 h to reach equilibrium prior to the initiation of the next current pulse. 2.4. Electrochemical stability of active material in aqueous electrolyte In order to investigate the stability and electrochemical activity of LCO samples in an aqueous electrolyte, open circuit potential testing and galvanostatic charge/discharge were also performed in a three-electrode cell. In these cells, the working electrode was made by pasting LCO slurry onto nickel foam current collectors (the slurry was prepared in the same way as for the coin cells described above). A platinum wire was used as the counter electrode and Ag/ AgCl electrode was used as the reference electrode. The electrolyte used in the cell was 1 M Li2SO4 aqueous electrolyte. All the electrochemical tests of the three-electrode cell were performed on a Gamry 1000 potentiostat. 3. Results and discussion 3.1. Phase and structure LCO has a rhombohedral symmetry with the space group R-3m [33]. The oxygen ions form close-packed planes stacked in an ABC sequence with cobalt and lithium ions occupying alternating layers of octahedral sites. This makes LCO suitable for accommodating large deviations in Li concentration [33]. Fig. 1 shows the XRD patterns for the four LCO samples. All of the strongest peaks for all samples were consistent with the LCO reference pattern, although at the lowest two calcination temperatures (750 C and 800 C) a small impurity peak consistent with Co3O4 was observed, as indicated with the * in Fig. 1 [34]. The ratio of the intensity of the (003) peak to the (104) peak in the XRD pattern, or the I (003)/I (104) ratio, was calculated for each of the LCO samples and these values are summarized in Table 1. The I (003)/I (104) ratio has previously been reported to correlate with the cation mixing between Liþ and Co3þ: a high I (003)/I (104) ratio indicates decreased cation mixing, or better segregation of Liþ in the Li layer and Co3þ in the Co layer in the LCO structure. Not only does Co3þ ions in the lithium layer

Fig. 1. XRD patterns for LCO900 (pink), LCO850 (green), LCO800 (blue), LCO750 (red), and the reference pattern for LCO (black, PDF 01-070-2685). The * marks the impurity peak consistent with Co3O4. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Table 1 Intensity ratio, I (003)/I (104), of the LCO samples. I (003)/I (104) LCO750 LCO800 LCO850 LCO900

1.75 1.73 1.97 2.70

decrease the available capacity, they also hinder Liþ diffusivity resulting in poor electrochemical performance, especially rate capability [35,36]. In these previous studies on layered oxides, the I (003)/I (104) ratio varied from 0.6 to 1.47. As we can see from Table 1, all of the synthesized LCO materials had I (003)/I (104) ratios > 1.7 which indicated relatively good segregation in their respective layers for the materials. However, LCO900 had the highest I (003)/I (104) ratio, which indicated the lowest level of cation mixing. In contrast, LCO750 and LCO800 had the lowest I (003)/I (104) ratios, which indicated significantly more cation mixing relative to the LCO materials fired to the higher temperatures and also suggested decreased electrochemical performance for these materials both with regards to absolute capacity and rate capability. Note the XRD analysis discussed above was on samples after the soft milling stage of processing. SEM was conducted on all the LCO samples to determine the morphology both before and after soft milling Figs. 2 and 3 show the SEM images of LCO samples before and after soft milling respectively. Before soft milling, all the LCO samples formed secondary particle clusters with sizes generally in the range of tens of micrometers and primary particles of ~couple hundred nanometers were loosely connected within the secondary particle clusters. As the calcination target temperature was increased, the primary particle size increased gradually due to the increased sintering from higher atomic mobility induced by the higher target temperature [5]. After soft milling, the secondary particle clusters were broken down into small clusters or individual particles. The primary reason for soft milling was to reduce the particle aggregate size to help mitigate sedimentation of particles within the suspension during DPR analysis.


825

Fig. 2. SEM images of synthesized LCO samples lithiated at a) 750 C, b) 800 C, c) 850 C, and d) 900 C. Images were taken before the subsequent soft milling step.

Fig. 3. SEM images of synthesized LCO samples lithiated at a) 750 C (“LCO750”), b) 800 C (“LCO800”), c) 850 C (“LCO850”), and d) 900 C (“LCO900”) after soft milling.

3.2. LCO suspension DPR characterization DPR analyses for the four LCO samples were conducted using the following procedure. 2 vol% LCO powders were mixed with 1 M Li2SO4 aqueous electrolyte to form particle suspensions. The suspensions were pumped through the custom flow cell using a pump at a rate of 80 ml min1. A series of chronoamperometry tests at sequentially increasing potentials were applied on the suspensions. The potentials were selected from 0.70 to 0.74 V vs. Ag/AgCl reference with an interval of 0.01 V. At each potential step, chronoamperometry was conducted for 60 s. The potential and average current after stabilization for each potential step in the chronoamperometry were retained for analysis. Applied potentials were plotted against the corresponding measured average currents,

and a linear regression was conducted on the data set from each DPR sequence. The slope of the linear fit of the potential as a function of the average current was a resistance, and this resistance was the DPR for a given sample. The typical timescale for a DPR measurement from adding the powder to the electrolyte to completing the last chronoamperometry step was around 20 min. An example of DPR analysis of 2 vol% LCO850 suspension is presented in Fig. 4a and the linear fit for determining the resistance is displayed in the figure inset. As potentials were increased, the plateau for the measured current increased accordingly, in a linear manner. The inset figure shows the resulting data obtained from the experiment in Fig. 4a, with applied potential vs. the average measured current of the last 10 s at that potential. For this DPR experiment, the slope was 850.0 U (the measured DPR), with an R2

826


Fig. 4. (a) Single DPR experiment (using LCO850) with inset showing applied voltage against average current of the last 10 s at that voltage. Line in inset represents best fit linear regression of the data. (b) DPR values for the different LCO samples at 2 vol% loading. The error bars represent three parallel DPR tests for each sample.

of 0.996, which indicated a good fit. Choosing aqueous electrolyte over organic electrolyte for DPR analysis was advantageous in several aspects: utilizing aqueous electrolyte improved the flexibility and cost of the technique (the analysis does not need to be in a glove box and aqueous electrolyte carrier fluid is much less expensive than organic electrolyte). Moreover, aqueous electrolyte has a much higher ionic conductivity than organic electrolyte (~100 mS cm1 for 1 M Li2SO4 aqueous electrolyte compared to ~15 mS cm1 for common LiPF6 in carbonates organic electrolyte [37]), which aids in reducing the bulk electrolyte resistance to a negligible value relative to the total measured DPR value. Although there are advantages of aqueous electrolyte relative to organic electrolyte in DPR analysis, the relatively narrow electrochemical window was expected to be problematic for cathode materials with high redox potentials, where the side reaction for water oxidation could contribute significantly to the measured current [38]. Consequently, galvanostatic charge/discharge tests and opencircuit potential (OCV) stability tests of LCO samples were conducted in a three-electrode cell in aqueous electrolyte prior to the DPR analysis. These experiments were conducted to determine the applied potential sequence during DPR measurements and to ensure that the LCO powders were stable in the aqueous electrolyte during the several minutes of DPR analysis. LCO samples were cycled in aqueous electrolyte at a rate of 0.2 C for 1 cycle (see Supporting Information, Fig. S3, for the charge/discharge curves). A discharge capacity of close to 100 mA h g-1 LCO was achieved. Note that this required the LCO electrode to cycle in the aqueous electrolyte for over 8 h continuously. The reversible discharge capacity suggests reasonable stability of the LCO in the aqueous electrolyte, even during electrochemical oxidation and reduction. Also, because the DPR measurements do not exceed 0.74 V (vs. Ag/AgCl), all of the DPR analysis was within the electrochemical stability window of the 1 M Li2SO4 aqueous electrolyte. The OCV stability test (see Supporting Information, Fig. S4) indicated that LCO powders were stable in the aqueous electrolyte for the duration of the 5 h of the test e because the OCV was stable for the measurement period. As mentioned previously, DPR was based on a concept of electrochemical reactions that occur only when solid particles in a dispersion were in contact with an electrode or current collector. The electrochemical response depends on an ensemble of particles that are electroactively connected to the current collector at a given time [26]. Accordingly, the DPR value of a sample can greatly vary with several factors including electrochemical cell design (cell size, current collector length and geometry), suspension flow rate, and active material loading; with all these factors being in addition to the intrinsic electrochemical properties of the active material [25]. Recognizing these variables, the cell design, suspension flow rate

and active materials loading were fixed for all the DPR experiments, and the resulting DPR values for the four samples are summarized in Fig. 4b. The DPR values of LCO samples decreased significantly as the target calcination temperature was increased. LCO750 had the highest DPR value of 1590.5 ± 168.8 U and LCO900 has the lowest DPR value of about 339.8 ± 21.4 U; LCO800 and LCO850 have DPR values of 1370.5 ± 96.2 U and 797.4 ± 55.6 U, respectively. 3.3. Electrochemical testing with conventional coin cells In order to compare the rate capability with DPR values, conventional coin cells were prepared to evaluate the rate capabilities of the four LCO samples. Galvanostatic discharge profiles at varying current rates (0.1 Ce5 C, 1 C corresponded to a current density of 137 mA g1) can be found in Fig. 5. For all cycles, the charge cycle was conducted at 0.1 C. The coin cells were cycled in the voltage range between 2.5 V and 4.2 V (vs. Li/Liþ). The full galvanostatic charge/discharge profiles can also be found in the Supporting Information, Fig. S2, for comparison of overpotential during cycling and energy efficiency of the electrodes containing the LCO materials. At low cycling rates, all the LCO samples have a discharge plateau at around 3.9 V. However, the initial discharge capacity at 0.1 C varies among the samples with LCO750 having the lowest capacity of 105 mA h g-1 and LCO900 having the highest capacity of 125 mA h g-1. With increasing discharge currents, the capacity retention decreased for all four LCO materials. The capacity retentions relative to the material's 0.1 C discharge capacity are shown in Fig. 6a. Again, LCO750 has the lowest capacity retention at higher discharge currents followed by LCO800. As for LCO850 and LCO900, there was no notable difference at currents of 2C and lower, but the capacity retention of LCO900 was much higher than that of LCO850 at 5C. The rate capability followed the order of LCO900 > LCO850 > LCO800 > LCO750. The results presented above demonstrated that DPR values indicated the relative rate capability of the materials e with a higher DPR indicating a lower rate capability. The exemplary correlation of LCO DPR values and discharge capacities can be found in Fig. 6b. DPR analysis was designed to only interrogate the intrinsic properties of the active materials, however, it is expected to be affected by several different variables, in particular, the material electronic conductivity, ionic conductivity, interfacial resistance, and particle size. All of these material properties would also have an impact on the rate capability of materials in a battery [18,20,39,40]. Based on the SEM and BET analysis (see Supporting Information S5), LCO750 had the smallest average particle size, which meant more total particles would be expected to participate in the electrochemical reaction at a given time during DPR analysis.


827

Fig. 5. Voltage profiles during galvanostatic discharge at increasing C-rates for (a) LCO750, (b) LCO800, (c) LCO850, and (d) LCO900. For all plots, the discharge curves correspond to 0.1C (black), 0.2C (red), 0.5C (blue), 1C (magenta), 2C (olive), and 5C (navy). All charge cycles were at 0.1C, and the voltage window used was 2.5e4.2 V (vs. Li/Liþ). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

And if the particles in contact with the electrode can be considered resistors connected in parallel, LCO750 would be expected to have the smallest DPR value based purely on these particle size considerations and the lower electron/ion transport distance for smaller sized particles [25]. However, we observed the contrary trend in the experiments e the largest particles with the lowest BET surface areas had the lowest DPR values. We speculated the DPR value must have been more strongly impacted by another intrinsic material property to reverse the expected trend of the particle size, and such material properties were expected to be either the material electronic or ionic conductivity. Compared with other common Li-ion battery cathode materials, LCO has a relatively high electronic conductivity, but the lithium ion diffusivity has been reported to vary greatly among LCO materials depending on the synthesis method [41]. While in future work we intend to explore the contributions of many different physical properties to different types of Li-ion materials' measured DPR values, here further insights were gained into the differences between the synthesized LCO materials by investigating the impact of the Liþ diffusivity in the LCO samples using GITT [44]. Lithium diffusivity can greatly affect the battery overpotential, particularly at high C-rates [33]. Therefore, high Liþ diffusivity in the crystal structure is important for a material to achieve good rate capability. We hypothesized that Liþ diffusion contributed to the measured DPR values. Thus, GITT analysis was conducted in coin cells as described in the experimental section. The GITT curves of the four LCO samples can be found in the Supporting Information, Fig. S6. The diffusivity of Liþ in the LCO crystal structure was calculated based on the following equation:

2 32 2 dE 2 IV M ~¼ 4 dx 5 t≪L D ~ dE p ZA FS D pffi 4

d t

(1)

~ is the Liþ ion diffusion coefficient, I is the current used in where D the titration, VM is the molar volume of lithium cobalt oxide, ZA is the number of charge of the ion species (in this case ZA ¼ 1 for Liþ), F is the Faraday constant, S is the electrolyte/electrode material contact area, x is the content of Li in the cathode materials (LixCoO2), t is the current pulse time and L is the characteristic pffiffi length dimension of the electrode material. The value of dE/d t is determined from the plot of the voltage response vs. the square root of the time during each current pulse, and dE=dx is obtained by plotting the equilibrium potential vs. the electrode material composition after each current pulse [42]. The electrolyte/electrode material contact area can be difficult to determine due to the introduction of the carbon conductive additives and polymer binders. Both the geometric and surface areas of the electrode determined by BET analysis have been used in the literature, which results in widely varying diffusion coefficients [33,44], though we note that other factors such as the detailed material and electrode conditions and the method of analysis (e.g., GITT, electrochemical impedance spectroscopy, cyclic voltammetry) will also influence the reported values. In this paper, the BET surface areas of the LCO powder were used to determine the contact area used in the GITT analysis. Fig. 7a shows a detailed view of LCO900 single current pulse with respect to time as well as its effect on the electrode equilibrium potential. Based upon the curve, the equilibrium potential before the current pulse was En. The current pulse induced bulk Liþ diffusion, and a new equilibrium potential will be formed after a long enough relaxation time. The equilibrium potential after the current pulse is Enþ1. The equilibrium potential change DE is calculated and used together with the known capacity extracted/ delivered during the current pulse to determine the value dE=dx, which was one of the inputs in Equation (1). Fig. 7b shows the detailed view of a single current pulse applied over time duration t. By plotting potential versus square root of time, a linear regression

828


Fig. 8. Liþ diffusion coefficients determined using GITT for LCO900 (pink diamonds), LCO 850 (blue triangles), LCO800 (red circles), and LCO750 (black squares), as a function of the extent of lithiation. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 6. (a)Discharge capacities of LCO900 (pink diamonds), LCO850 (blue triangles), LCO800 (red circles), and LCO750 (black squares) at different C-rates relative to the capacity at 0.1C. Lines added to guide the eye. (b) DPR values of LCO samples and corresponding discharge capacity retention at 5C. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

pffiffi was conducted in order to calculate dE/d t term in the diffusion coefficient calculation Equation (1). Note that the diffusivity calculation from GITT is based on Fick's law, which dictates that the potential is linear with respect to the square root of time when the process is only limited by transport via ion diffusion [43]. As shown in Fig. 7b, the linear fit R2 value was 0.988, indicating a good fit. The slope obtained from the fit was used in Equation (1) to determine

diffusivity. Fig. 8 shows the calculated diffusion coefficients for the four LCO samples during discharge at different extents of lithiation of the LCO materials. The lithium compositional range was divided into the solid-solution reaction region and the two-phase region. As mentioned previously, the GITT analysis was only appropriate in the solid-solution reaction region due to its derivation from Fick's law [45]. The diffusion coefficient calculated in the two-phase region was not only affected by lithium ion diffusion, but also by the energy barrier imposed by the phase transformation. By comparing the lithium diffusivity for the four LCO samples in the solid-solution region, it was observed that lithium diffusivity from high to low followed the order of LCO900 > LCO850 > LCO800 > LCO750. At a composition of Li0.7CoO2, LCO900 had the highest lithium diffusivity at 1013.8 cm2 s1 while LCO750 has the lowest lithium diffusivity at 1014.3 cm2 s1. These measured diffusion coefficients were consistent with the I (003)/I (104) ratio analysis because the material with higher cation ordering would be expected to have a higher Liþ diffusion coefficient. These measured values for the diffusion coefficients were comparable to previous reports from the literature for LCO [33,44,46]. Lithium diffusivity during the charging process was also calculated for all the samples and can be

Fig. 7. (a) Detailed view of single current pulse during charging of the cell using the GITT analysis, with relaxation potentials before (En) and after (Enþ1). The time of the current pulse, t, was 600 s. (b) Detailed view of a plot of potential as a function of square root of time used during GITT analysis. The red line represents a plot of the linear line of best fit after least squares regression. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)


found in the Supporting Information, Fig. S7. The diffusion coefficients for the charging process followed the same trend as the discharge process. The variance of Liþ diffusivity in the crystal structure among these four LCO samples was in good agreement with the DPR values, and also aided in rationalizing the different rate capability performance of the four LCO samples. The results presented above were consistent with the hypothesis that the ionic conductivity of the LCO material was the major contributor to the differences in the measured DPR. We note here that it is expected that the relationship between DPR and relative rate capability would be relatable to battery cells and electrodes under conditions where the rate capability was dependent on the material properties of the active material, such as the Liþ diffusivity, electronic conductivity, or active material particle size. However, under conditions such as very high rate charge/ discharge or very thick electrodes, other factors can limit the capacity retained at high rate such as the electrolyte ionic conductivity, the electronic conductivity of the electrode matrix, and the transport of ions through the electrode matrix. When rate capability is limited by these factors DPR would not be expected to provide a direct relative comparison to rate capability, although the technique would still provide insights into the active material used in such cells. Also, to appropriately normalize the measured DPR the amount of the active material in contact with the current collector needs to be reliably estimated, and techniques to determine these estimates will be the subject of future research. In addition, the material in this DPR investigation was only being probed when it was pristine and during the electrochemical oxidation process, and the ability to run DPR at different states of charge and both while oxidizing and reducing the electroactive particles would provide broader insights into the material properties. 4. Conclusions DPR analysis was performed on four LCO samples in an aqueous electrolyte. DPR values obtained from the analysis were shown to correlate with and be predictive of the rate capability of the LCO samples, with the rate capability being confirmed via conventional coin cell cycling. DPR was a relatively fast (