Quantitative analysis of Li contents in Li x CoO 2 cathodes via Rietveld refinement

Transcription

1 Quantitative analysis of Li contents in Li x CoO 2 cathodes via Rietveld refinement Mark A. Rodriguez, a) David Ingersoll, and Daniel H. Doughty Sandia National Laboratories, PO Box 5800, MS 1411, Albuquerque, New Mexico Received 13 May 2002; accepted 17 October 2002 Rietveld refinement of Li x CoO 2 -type cathodes where x ) has been demonstrated to yield quantitative information about the Li occupancy with an error of about 10%. With careful X-ray diffraction XRD data collection, refinement, and proper calibration, accurate values can be obtained. Rietveld refinement tends to under-predict the Li occupancy values in charged de-lithiated cathodes as compared to ICP measurements. A Li-gradient model that assumes a decreasing concentration of Li from the particle core to its surface is discussed to explain this observation. The observed lattice parameters for the fully lithiated LiCoO 2 standard phase R-3m symmetry were a (1) Å, c (1) Å. The error convention used in this paper is 3 of standard deviations obtained from Rietveld refinement output International Centre for Diffraction Data. DOI: / Key words: Rietveld, Li x CoO 2, lithium, cobalt, oxide, Li-ion battery INTRODUCTION Li-ion batteries are a promising technology for rechargeable power supplies. Many consumer electronics such as camcorders, laptop computers, and cameras already employ these rechargeable batteries. There are several benefits for utilizing these batteries. They offer higher energy density than the Ni Cd or Ni H technologies, are lightweight, and function at higher operating voltages. These factors have stimulated a great deal of research and development in Li-ion technology Jacoby, Although these batteries show promise, significant challenges remain. Issues concerning cathode and anode stability, capacity optimization, and battery integration all require further research. More importantly, a solid fundamental science-based understanding of the battery performance is critical to identify failure mechanisms, thereby establishing methods for battery improvement. This fundamental understanding was the motivation of our research into characterization of Li-ion cathode materials. Li-ion batteries are often referred to as rocking-chair batteries Doughty, 1996; Koksbang et al., 1996 since charging and discharging of the battery can be thought of in terms of the intercalation and de-intercalation of Li between a host cathode in this case LiCoO 2 ) and anode graphite. As Li ions are shuttled from anode to cathode, electrons traverse an external circuit used to power a device. Hence, in analyzing Li-ion batteries, the location of lithium is critical for understanding the state of charge and condition of the battery. Unfortunately, due to the low number of electrons, Li occupancy is notoriously hard to detect via X-ray diffraction XRD. Structural analysis of powders is typically performed using neutron diffraction Lee and Kim, 2000 ; however, neutron diffraction is not a routine technique. Access to neutron diffraction facilities is highly competitive and typically only a few samples can be characterized. Additionally, sample sizes need to be quite large, usually on the order of a Electronic mail: marodri@sandia.gov several grams of powder. All of these drawbacks make neutron diffraction less appealing as a method of structural characterization. Therefore, it was thought that revisiting the idea of structural analysis via XRD might prove fruitful for quantitative analysis of the Li content in cathodes. Even if the results of Li site occupancy had large errors, if there was any useful information about Li contents that could be correlated to performance, it would be of interest as a possible characterization technique. Additional benefits of XRD analysis include a relatively fast turnaround, easy access, the ability to analyze small sample sizes, a relatively low cost, and the fact that it is a nondestructive technique. We present results of a systematic study of Li content in commercial Li-ion batteries via X-ray Rietveld refinement and ICP analysis to determine the feasibility of employing XRD as a method of Li content determination. EXPERIMENTAL Sample preparation Sony Li-ion batteries were charged to various known voltages and subsequently disassembled. The partially de-lithiated Li x CoO 2 cathode was removed from the battery in a dry-box, rinsed with an anhydrous DEC/DMC solvent and dried under vacuum in the anti-chamber of the glove box. Since the as-deposited cathode had significant c axis orientation on the aluminum current collector, the cathode powder was detached from the current collector and ground in a mortar and pestle for 45 minutes to obtain a randomized powder specimen. XRD data collection Specimens were mounted on zero-background holders using a top-drifted technique Jenkins and Snyder, Special care was taken to assure that the specimen covered the entire X-ray beam footprint on the sample holder. Additional care was taken to avoid preferred orientation effects. 135 Powder Diffraction 18 (2), June /2003/18(2)/135/5/$ JCPDS-ICDD 135

2 First, the sample was examined under a metallograph to assure that there were no unusually large, oriented particles. Second, a fast scan of the diffraction pattern was collected and quick refinement of the March value March, 1932 was determined. A criterion was placed on the specimen that if the March value dropped below 0.9 (1.0 complete randomization, the sample would be prepared again. This typically did not happen since all samples prepared using the method described above met the March criterion and were subsequently analyzed. Diffraction patterns were collected on a Siemens diffractometer equipped with a standard sealed tube X-ray source Cu, a diffracted beam monochromator, and a scintillation detector. Parameters for typical diffraction scans were range, 0.04 step-size and a counttime of 20 seconds. The generator values used were 45 kv and 30 ma. ICP data collection The cathode samples were analyzed for cobalt Co, nickel Ni, and lithium Li by inductively coupled plasmaatomic emission spectroscopy ICP-AES. Each sample was dissolved in nitric acid and gently heated on a hot plate until dissolution was complete. A Perkin Elmer Optima 3000 ICP- AES spectrometer system was used to analyze the solutions. Quantitative analysis was performed with calibration curves generated using aqueous standards prepared from NISTtraceable Ni, Co, and Li standard solutions. Each solution was analyzed in triplicate on the ICP-AES instrument. The result for each sample was reported as the average of the nine measured values triplicate ICP-AES measurements on triplicate sample preparations. Estimated uncertainties were reported as 95% confidence limits, which represent fit of standard calibration curves, instrument precision, and triplicate sample preparation. SEM analysis of powder Scanning electron microscopy SEM was performed on cathode powders to determine the particle size. SEM samples were prepared by distributing ground cathode powder onto carbon tape attached to aluminum sample stubs. Micrographs were recorded using a Hitachi S4500 field emission gun FEG, high resolution SEM. RESULTS AND DISCUSSION Calculated patterns Initial interest in using XRD data to investigate Li content was based on calculated patterns. Calculated powder patterns generated by the program Ruby Materials Data, Inc. were being employed for comparison to observed powder patterns. In the process of comparing these patterns, it become clear that certain reflections were more dependant on the Li content than others. This is clearly illustrated in Figure 1. This figure shows two diffraction patterns. The thin solid line is the calculated pattern for Li x CoO 2 (x 1) or fully lithiated. The thick solid line shows how the diffraction pattern changes if the Li occupancy is reduced to x 0 while all other variables lattice parameters, oxygen position, thermal Figure 1. Calculated patterns for Li x CoO 2 showing intensity variation with Li removal. parameters, etc. were held constant. The patterns were scaled with the 003 peak, not shown, set as the 100% peak of the pattern. Although the x 0 is not a true structure in that there will surely be changes in the other variables as the cathode de-lithiates as well as structural instability as the Li approaches x 0, this calculation shows us the general impact of Li removal on the diffraction pattern. For instance, it is clear from the comparison of these patterns that certain hkl s are affected by the Li site occupancy, while others are not affected at all. The most dramatic effect is seen for the 104 reflection where the peak shrinks as much as 25% as x is reduced from 1 to 0. Other reflections were affected too, such as the 006 and 012 peaks. However, peaks such as the 101 and 105 as well as the 003 which was not shown in the figure, were not affected by the reduced Li site occupancy. Because of the asymmetric nature of intensity variation due to Li site occupancy and because the largest change was seen in the 104 peak, the second most intense peak of the LiCoO 2 diffraction pattern, it was considered likely that structure refinement of XRD data could indicate something about Li site occupancy. Therefore, Rietveld refinement was attempted on these cathode materials. Rietveld refinement of Li x CoO 2 cathodes Rietveld structure refinements of the Li x CoO 2 cathodes were carried out using the program RIQAS Materials Data, Inc.. The same rhombohedral/hexagonal structure model Lee and Kim, 2000 was used for all data sets. It has been shown that pure LiCoO 2 undergoes a monoclinic distortion during Li removal Reimers and Dahn, We did not observe this behavior in our samples. In our ICP measurements a small amount of Ni ( 0.2 wt%) was detected. Perhaps the presence of this Ni dopant stabilized the hexagonal lattice, or perhaps a monoclinic structural distortion of the unit cell was so small that it did not significantly impact the refinement. Whatever the reason, the hexagonal-model fit the data well and was used to fit all Li x CoO 2 XRD patterns. A total of 17 parameters including five structural parameters were refined. Since Li and Co both occupy unique sites in this R-3m structure, only the oxygen atom occupies a site with an adjustable positional parameter Lee and Kim, 136 Powder Diffr., Vol. 18, No. 2, June 2003 Rodriguez, Ingersoll, and Doughty 136

3 TABLE I. Rietveld refinement results for Li x CoO 2 cathodes compared with Li content via ICP. Sample ICP Li Li occ Co occ a axis Å c axis Å Vol Å 3 Oz R 1 FMC STD Sony 3.50V Sony 4.04V Sony 3.80V Sony 3.98V Sony 3.84V Sony 4.10V Sony 4.15V Sony 4.20V Hence, the 5 structural parameters refined were the a and c axes, Li site occupancy, Co site occupancy, and Oz shift of the oxygen along the c axis. Temperature parameters for atoms were fixed at B iso values obtained from neutron diffraction measurements Lee and Kim, 2000 where Li 1.2 Å 2,Co 0.16 Å 2,O 0.48 Å 2. The atomic scattering factors for Li, Co, and O were taken from the International Table for X-ray Crystallography IV Ibers and Hamilton, Crystal data for the Rietveld refinements are shown in Table I, along with Li contents from ICP measurements for comparison purposes. Errors reported in Table I are 3 times greater than the output from the Rietveld refinement program and hence indicate a 3 or 99% confidence interval for the refined values based on the Rietveld fitting routine. This was done to add credibility to the values and not over-interpret the data. Additionally, a fully lithiated LiCoO 2 powder obtained from FMC Corporation, Philadelphia, PA was also measured to establish credibility for the diffraction and ICP measurements. This sample shall be referred to as the FMC standard throughout this manuscript. Based on the data in Table I, it appears that even with the conservative restriction of 3 placed on the Li site occupancies observed from the refinements, the Li occupancy could still be established within an error of about 10%. The data in Table I were placed in order of decreasing Li content based on ICP measurements. When the data are placed in this order all the structural parameters for the Li x CoO 2 cathodes show clear trends e.g., consistent unit cell volume increase with decreased Li content. It was no surprise that the FMC standard showed the highest Li content at x 1) and the XRD and ICP are in very good agreement with this expected result of a fully lithiated powder. Hence, it appears that both ICP and XRD techniques are measuring valid Li contents in these cathode materials. Additionally, the residual error in the refinements (R 1 ) appear reasonable for a good fit of observed and calculated data, giving additional validity to the refined structural parameters. Interestingly enough, the Li ICP contents track somewhat with battery voltage as indicated in the sample label, however, this trend does not always hold. For example, the Sony 4.04V cathode best fits between the Sony 3.50V and 3.80V cathodes. Likewise, the Sony 3.98V cathode best fits between the 3.80V and 3.84V cathodes. Clearly the battery voltage is not a trustworthy method of tracking Li content. This can be explained by the possibility of battery shorting that can occur after charging. The shorting can cause a more highly charged battery to re-intercalate Li into the cathode and hence increase the observed Li content in the cathode. Therefore, one can already see the sensitivity of this analysis for identifying shorting effects in Li-ion batteries. Additional evidence of the voltage shorting is demonstrated by the bond length data shown in Table II. When the results are arranged in a similar manner to Table I i.e., decreasing Li content based on ICP, the bond lengths show a clear trend toward contracting Co O bonds and expanding Li O bonds as Li is removed from the structure. This behavior makes sense as it reflects the increased oxidation of cobalt as well as the loosening of the Li O bond as Li is extracted from the Li layer. There are some inconsistencies in the bond length trends; most notable are the observed values for the 4.04V cathode. It must be pointed out that these bond lengths are almost entirely dependant on the refined Oz position. Hence the accuracy of this atom position will dictate the quality of the bond lengths. When all the data are taken together, there is a clear correlation between ICP Li content and bond length. As one considers the Li contents, one observes that the XRD measurements decrease along with the ICP measurements. This can be illustrated graphically by plotting the refined Li occupancy via Rietveld vs ICP Li content as shown in Figure 2. If there was a prefect agreement between the two measurements, the data should generate a linear relationship with all the data points falling on the straight line indicated in the plot as the ideal relationship. The actual data do not all fall on this ideal line. Only the first data point for the fully lithiated FMC standard displays the ideal behavior. All the other data points fall on a linear relationship that deviates from the ideal curve. The data do not seem to suggest that there is some systematic calibration error that causes this deviation, since that effect would more likely result in the proper slope of the observed data while requiring some constant shift of the trend-line up or down to obtain TABLE II. Bond lengths for Li x CoO 2 cathodes based on Rietveld refinements. Sample ICP Li Li O bond Å Co O bond Å FMC STD Sony 3.50V Sony 4.04V Sony 3.80V Sony 3.98V Sony 3.84V Sony 4.10V Sony 4.15V Sony 4.20V Powder Diffr., Vol. 18, No. 2, June 2003 Quantitative analysis of Li contents in Li x CoO 2 cathodes

4 Figure 4. Schematic model of Li concentration within a Li x CoO 2 particle. Figure 2. Comparison of observed Li contents based on XRD Rietveld and ICP measurements. the ideal type behavior. The data actually suggest that, as the cathode becomes more and more de-lithiated, the Rietveld refinement of Li occupancy increasingly underestimates the true Li content as observed by ICP. This observation appears to be real and will be discussed below as a result of Ligradients present in the cathode powder. Lithium gradient model It is reasonable to assume that during de-lithiation of the cathode, the Li on the outside of the LiCoO 2 particles will be removed first. Continued de-lithiation will require extraction of Li closer and closer to the core of the particle. Figure 3 shows an SEM micrograph of a typical cathode powder sample. Measurements of particle size from the SEM micrographs yielded an average value of 12 m for Li x CoO 2 grains. A schematic illustration of a Li x CoO 2 particle is shown in Figure 4. In this figure the particle has had a significant amount of Li extracted. This particle contains a Lirich core as well as a variation of Li concentration from the core out to the surface of the grain. A penetration depth calculation for LiCoO 2 which is almost completely a function of Co content shows that 50% of the information obtained in the X-ray diffraction pattern results from the first micron of the particle. Likewise, 75% of the data comes from the first three microns; only a small fraction comes from the interior core of the particle. Therefore, the XRD measurement and refinement will bias the data to the surface of the cathode particle. ICP measurements, however, dissolve the powder completely prior to analysis and thereby obtain a much more accurate measure of the full Li concentration since the Li-rich core is also measured. This Li-gradient model explains the observed results illustrated in Figure 4. Although the XRD measurement fails to predict the observed Li content as shown by ICP, the relationship between XRD and ICP Li contents is linear and can be explained by the Li-gradient model. This means that such calibration curves can be generated for Li content and that additional XRD measurements can be corrected based on the calibration curve, to obtain a quantitative Li content for cathode materials in a relatively fast, nondestructive manner. Additionally, even though the XRD analysis may bias the data as a particle-surface measurement of Li content, this obtained Li concentration may be a more valuable measurement for correlation to performance of the battery than the more accurate value obtained by ICP. This is because possible failure mechanisms may manifest themselves in the more strongly de-lithiated Li x CoO 2 which exists at or near the surface of the particle. So even though the overall Li content of a particle may be, say x 0.3, perhaps the outside surface has been reduced to as low as x 0.1, resulting in a composition that no longer displays good cycling and ultimately leads to loss of capacity and cycle life. Hence, both the uncorrected and corrected Li concentrations will be of some value for research and development of Li-ion batteries. CONCLUSION Rietveld refinement of Li x CoO 2 -type cathodes can yield quantitative information about the Li occupancy with an error of about 10%. Care in data collection and proper calibrations are necessary to obtain accurate values. The Rietveld refinement tends to under-predict Li occupancy values in charged de-lithiated cathodes. A Li-gradient model that assumes a decreasing concentration of Li from the core of a particle to the surface can explain this observation. Figure 3. SEM image of LiCoO 2 30 microns. powder on carbon tape scale length ACKNOWLEDGMENTS The authors would like to thank Jill Langendorf and Lorie Davis for their help with cathode preparation. The authors 138 Powder Diffr., Vol. 18, No. 2, June 2003 Rodriguez, Ingersoll, and Doughty 138

5 also thank Michael Kelly and Jeanne Barrera for ICP analysis. Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy under Contract No. DE-AC04-94AL Doughty, D. H Materials issues in lithium ion rechargeable battery technology, SAMPE journal 32, Ibers, J. A. and Hamilton, W. C International Tables for X-ray Crystallography, Vol. IV Kluwer Academic, Boston, p.71. Jacoby, M Taking charge of the 21st century, Chem. Eng. News 76 31, Jenkins, R. and Snyder, R. L Introduction to X-Ray Powder Diffractometery Wiley, New York, pp Koksbang, R. et al Cathode materials for lithium rocking chair batteries, Solid State Ionics 84, Lee, K. K. and Kim, K. B Electrochemical and structural characterization of LiNi 1 y Co y O 2 (0 y 0.2) positive electrodes during initial cycling, J. Electrochem. Soc. 147, March, A Mathematische theorie der regelung nach der Korngestalt bei affiner deformation, Z. Kristallogr. 81, Reimers, J. N. and Dahn, J. R Electrochemical and insitu x-ray diffraction studies of lithium intercalation in Li x CoO 2, J. Electrochem. Soc. 139, Powder Diffr., Vol. 18, No. 2, June 2003 Quantitative analysis of Li contents in Li x CoO 2 cathodes