Hye Jin
Kim‡
a,
Youngkyu
Park‡
a,
Yoonjin
Kwon
a,
Jaeho
Shin
a,
Young-Han
Kim
b,
Hyun-Seok
Ahn
b,
Rachid
Yazami
c and
Jang Wook
Choi
*a
aSchool of Chemical and Biological Engineering and Institute of Chemical Processes, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Republic of Korea. E-mail: jangwookchoi@snu.ac.kr
bKorea Electronics Technology Institute (KETI), 11 World Cup buk-ro 54-gil, Mapo-gu, Seoul 03924, Republic of Korea
cKVI PTE LTD, 2 Cleantech Loop, Singapore 637144
First published on 4th December 2019
Upon the emergence of electric vehicles, accurate and non-destructive monitoring of battery electrode materials during operation is highly desirable. Structural degradation of widely adopted intercalation-based materials constitutes the origin of their capacity fading and safety deterioration. Here, we introduce entropymetry to monitor the structural changes of LiCoO2 (LCO) and its nickel (Ni)-doped derivatives at different states of charge (SOC). While simple lithium (Li) extraction on charging gives a monotonic decline of entropy change (ΔS) based on progressive vacancy occupation over Li sites, the presence of a monoclinic intermediate phase inverses the slope of the ΔS profile to reflect its limited atomic configurations with high ordering. Furthermore, Ni-doping lessens the ordering of the monoclinic phase, decreasing the height amplitude of ΔS profile in the monoclinic regime. The increased disorder by Ni-doping enhances the stability of the lattice framework, extending the cycle life with high voltage cut-off (4.6 V vs. Li/Li+). The present study highlights entropymetry as a unique, non-destructive tool in monitoring the structural ordering and relevant degradation of electrode materials in lithium-ion batteries.
Broader contextThe required cycle life of rechargeable batteries for electric vehicles (i.e., 80% capacity retention for 10 years over 100000 miles) is a daunting challenge, even for state-of-the-art battery systems. While champion materials and optimized cell conditions are essential for this purpose, intelligent cell operation is equally important in extending cell lifespan. Although battery management systems serve such a role, their functions are usually based on current, voltage, and temperature, without taking the chemical state of electrode materials into account. In this sense, the current study suggests entropymetry as a useful diagnostic tool in monitoring the state of electrode materials in a non-destructive fashion. The lattice configuration alteration involving Li and transition metal changes the entropy of the material, offering useful structural information of LiCoO2, a model material of interest in this study. With the same logic, acquisition of entropy change can capture the lattice ordering change of an intermediate phase, such as the monoclinic phase in LiCoO2. Different amounts of Ni-doping to LiCoO2 were detected by the entropy change in the monoclinic intermediate phase and correlated with improved cycle life via a stabilized oxygen framework bearing lithium–nickel cation exchange. Hence, entropy change can serve as a key parameter for non-destructive diagnosis of the state and evolution of a battery electrode material, which can provide useful input for an intelligent operation algorithm. |
Many analytical tools based on various microscopes,6–9 spectrometers,10–14 and diffractometers15–17 are currently used to characterize the crystal structures and atomic bond characters of battery electrode materials. However, most of these analyses require cell disassembly, along with high-cost equipment, and are thus ‘destructive.’ Not only is it a tedious task to recover the cell components for post-mortem analysis, but it also renders the cell useless, which is undoubtedly cost-ineffective. Therefore, a simple yet non-destructive analytic method by which the states of battery materials can accurately be diagnosed during operation would certainly save time and cost.
In this respect, entropy is a valuable parameter for gauging the current health status of a battery. Entropy is a state function related to the degree of disorder or randomness of a system. From a statistical mechanics viewpoint, entropy refers to the number of arrangements of atoms in a system of atomic array. In particular, for LIB electrode materials where Li+ ions are reversibly (de)intercalated (from)into the host structure during cycling, entropy is associated with the number of possible lattice configurations in which Li atoms are arranged with other lattice atoms.18,19 According to the Boltzmann definition of entropy (S),
S = kB·lnW | (1) |
As the crystal structure continuously changes and degrades over cycling, the entropy profiles would also change. Therefore, entropy change (ΔS) measurements, so-called entropymetry, can provide valuable information about structural transformations of electrode materials.
ΔS can be measured based on the following rationale. Assuming a quasi-equilibrium state where an infinitesimal amount of Li+ (Δx) is extracted from the cathode, the change of Gibbs free energy (ΔG) is equal to the electrical work (We) done at the open-circuit voltage (OCV):
ΔG = ΔxFE = We | (2) |
Dividing both sides by Δx and approximating it to zero,
(3) |
From Maxwell relations at the isobaric condition,
dG = VdP − SdT = −SdT | (4) |
(5) |
Combination of the eqn (3) and (5) results in:
(6) |
Therefore, ΔS of electrode materials can be assessed by measuring OCV change over varying temperatures.
There have been some attempts to measure ΔS of a battery cell,20–27 but an in-depth correlation of ΔS to structural changes was not provided. In fact, to date, entropymetry has not been seriously considered as a reliable diagnosis tool for battery electrode materials during cell operation. In the present work, we demonstrate entropymetry as a non-destructive analytical technique to monitor the structural changes of LiCoO2 (LCO) and nickel (Ni)-doped LCO during their battery cycling. Although diverse layered metal oxide analogues are adopted or under consideration as cathode materials for commercial lithium-ion batteries, LCO is still the most widely adopted, especially for mobile IT devices due to extensive knowledge of the correlation between its structural properties and long-term cyclability, along with a well-established manufacturing process. However, LCO suffers from unwanted phase transitions and structure destabilization upon exposure to high voltage operations (i.e., >4.5 V vs. Li/Li+).28,29 In order to enhance the structural stability of LCO at high voltage, several studies were carried out to replace cobalt (Co) with other metals such as Al, Ni, Mg, etc.30–34 Benchmarking those approaches, we synthesized LCO and Ni-doped LCO via a sol–gel process and investigated the effect of Ni-doping on structural stability over cycling. More importantly, we identified the structural evolutions of those materials by following ΔS and demonstrated entropymetry as a useful, non-destructive tool in analyzing battery electrode materials during cell operation.
Fig. 2 Comparison of entropy and ΔS profiles in the cases (a) without and (b) with the interaction between Li+ ions. |
The original structure of fully-lithiated LCO retains the hexagonal phase with O3 stacking, which is denoted as O3(I). As Li+ de-intercalates, the repulsive force between O2− in the framework increases, resulting in a gradual lattice expansion along the c-axis and a phase transition from O3(I) to O3(II).36,37 The newly formed O3(II) phase has the same hexagonal crystal structure as the original O3(I) with a larger c-axis lattice constant. The flat ΔS plateau is attributed to the co-existence of the two phases (O3(I) + O3(II)) forming phase boundaries. According to the Gibbs phase rule, when two phases are present in a binary component system (Li and the host) at constant total pressure and temperature, an additional degree of freedom is unavailable. This equally represents an independent ΔS from the Li composition, resulting in a flat ΔS profile with respect to composition change.
Another region of interest (0.4 < x < 0.46) is where ΔS increases sharply, which appears to be associated with Li-vacancy ordering as described above. Previous theoretical and electron diffraction studies also explained this Li-vacancy ordering phenomenon based on Li+–Li+ repulsion.38–40 Such Li-vacancy ordering leads to distortion of the layered structure, giving rise to monoclinic symmetry. The decreasing slopes of the ΔS profile in other regions can be explained by gradual reduction of Li content as described in Fig. 2a. In this way, the ΔS profile in each region is well correlated with the atomic configuration and the structural changes, and the ΔS profile follows the same trend with the voltage profile in terms of slope variation (Fig. S1, ESI†).41,42 Furthermore, the entropy change profiles of LCO were reproducible with multiple cells (Fig. S2, ESI†) and were also consistent with those in the literature.18
Fig. 4 Schematic diagrams of phase transition during Li+ extraction from (a) LCO and (b) Ni-doped LCO. |
In situ XRD analysis was performed to examine the effect of cation mixing on the crystal structure. Fig. 5a shows the integrated intensity ratios of I(003)/I(104) for each sample at the pristine state. This intensity ratio reflects the degree of the cation mixing, and decreases from 1.77 to 1.46 as the Ni content increases from 0 at% to 5 at%. After pre-cycling, in situ XRD analysis was carried out while the three electrodes were charged. As displayed in Fig. 5b, the charging profiles of those electrodes are quite similar with almost the same specific capacities regardless of Ni content. In Fig. 5c–e, the progressive XRD profiles reveal that the (003) peaks of the three electrodes are similarly shifted to the left and then to the right throughout charging. Also, the three electrodes undergo consistent sequential phase transitions from O3(I) to O3(II) then to H1-3. As shown in Fig. 5f, however, the leap in the c-axis distance corresponding to the O3(I) to O3(II) transition is delayed as the amount of Ni substitution increases, indicating a latent formation of the O3(II) phase. Moreover, a sharp plunge of the lattice constant related to the H1-3 phase formation that occurs above 4.55 V follows distinct behavior for the three electrodes. The extraction of Li+ at this high voltage regime triggers a slippage of the metal-oxide stacks and even Co ion migration to the Li layer, causing irreversible spinel phase formation such as LixCo2O4 and Co3O4.45 Thus, the delayed formation of the H1-3 phase upon Ni incorporation indicates its effect toward enhanced structural stability at high operation voltages.
The changes in the ΔS profiles before and after the cycling are plotted for Co100, Co97, and Co95 (Fig. 7b–d) in order to check if the initial cation mixing induced by Ni-doping is the primary reason for the superior cycling performance. Although the overall shapes of the ΔS profiles do not change much for all samples after 50 cycles, the change of trough-to-peak height in the monoclinic region of 4.05–4.2 V is different in a way that the height at the precycled state is the greatest for Co100 and declines with increasing the amount of Ni-doping. The trough-to-peak height differences before and after the cycling are 43.9, 32.5, and 16.7 for Co100, Co97, and Co95, respectively (Table S2, ESI†). Based on the logic described above, the decrease in the trough-to-peak height represents that the ordering has weakened as the cycle progresses. These results, in correlation with the cycling performance, indicate that initially lowered ordering of an intermediate phase by positioning transition metals in the Li layer can enhance high voltage structural stability by reducing internal strains.
To further correlate ΔS behavior with the structural ordering in a quantitative manner, the trough-to-peak height is plotted versus I(003)/I(104) from XRD analysis as shown in Fig. 7e. In the case of the pristine electrodes (marked with dots) that did not go through cycling, their trough-to-peak heights show a notable trend related to the degree of cation mixing; as I(003)/I(104) ratio decreases, the trough-to-peak height decreases. However, it is noted that the I(003)/I(104) ratios of the cycled electrodes are not suitable to quantify the degree of cation mixing because cycling causes particles in the planar electrode to become aligned along the (00l) direction irrespective of phase transition.47 Therefore, the I(003)/I(104) ratios of the cycled samples are linearly interpolated from the entropy change plots of the pristine samples, instead of directly plotting based on their actual I(003)/I(104) values. The inferred I(003)/I(104) ratios of the cycled samples (marked as stars) decrease compared to those before cycling, suggesting increased cation mixing during cycling. The trough-to-peak heights of the cycled Co100 and Co97 are both positioned between those of the pristine Co97 and Co95. To confirm the validity of this trend, pristine Co93 was additionally prepared and subjected to the same entropy analysis (Fig. S4, ESI†). Following the same trend, the trough-to-peak height of Co93 is lower than that of the pristine Co95, and the value of the cycled Co95 is between those of the pristine Co95 and pristine Co93. When I(003)/I(104) ratio is compared before and after 50 cycles, the difference before and after the cycling becomes larger with higher Co content (Fig. 7f); the I(003)/I(104) ratio differences before and after the cycling are 0.21, 0.13, and 0.07, respectively, for Co100, Co97, and Co95. The trend of trough-to-peak heights obtained from the series of samples and their cycled derivatives, linked with the I(003)/I(104) ratio changes, suggests that trough-to-peak height can serve as an indicator of the degree of cation mixing. Hence, the structural evolution of Ni-doped LCO can be non-destructively monitored during its lifetime via ΔS landscape. Furthermore, ΔS information can be a crucial component in detecting cycle life and safety weakness as these are heavily related to cation mixing.48,49 Thus, ΔS information can synergistically complement pre-existing lifetime and safety prediction models.50
The XRD analysis of the Co100, Co97, and Co95 electrodes before and after high voltage cycles indicates that the layered structures are well maintained and no additional phases were observed during cycling (Fig. 8a). However, the (003) peak position after cycling changes as the Ni content is varied. In the general case of layered metal oxide cathodes, over-potential during discharge inhibits the complete re-intercalation of Li ions, which leaves the c-axis lattice constant extended due to oxygen–oxygen repulsion between the oxygen-based lattice stacks.51 Such stretched c-axis is usually observed by a leftward (003) peak shift to a lower 2-theta value. On the contrary, the (003) peak of the cycled Co100 shifts to a higher 2-theta value compared to that of its pristine counterpart. This result indicates a decreased c-axis lattice constant due to irreversible structural deterioration by the slippage of metal oxide slabs at high voltages.15 The (003) peaks of the cycled Co97 and Co95 remain at lower 2-theta values than that of the cycled Co100. This trend in the peak shift reveals that Ni-doping makes the phase transition of LCO during high voltage cycling more reversible and suppresses unwanted structural changes. Hence, these XRD results are commensurate with the above ΔS results in that the ordering decreases more significantly with greater Co content.
Raman spectroscopy was employed to investigate the local structural disorder related to the phase transformation. Fig. 8b shows the Raman spectra of the Co100, Co97, and Co95 electrodes before and after 50 cycles. The Raman spectra of pristine LCO-based samples show two main peaks at around 486 and 585 cm−1 corresponding to Eg and A1g active Raman modes, respectively.52,53 In addition, a weak peak at 665 cm−1 was detected, which can be assigned to spinel phases such as Co3O4 and LixCo2O4 originating from the loss of Li during synthesis.54 After cycling, all three electrodes show relatively stronger spinel phase peaks at around 505 and 665 cm−1, representing irreversible phase transition to spinel phases during cycling. The intensity of the peaks assigned to spinel phases (Co3O4 and LixCo2O4) is the highest for Co100, which can be explained by the fact that the cation mixing of the Ni-doped samples suppresses the structural degradation during cycling. Therefore, the Raman spectra portray a consistent picture with the entropy analysis such that Ni-doping contributes to maintaining the lattice framework against severe irreversible structural transformation involving the electrostatic repulsion of the oxygen lattice.
Fig. 8c displays the Co 2p XPS spectra of the three electrodes before and after the high voltage cycling. As shown in Fig. 8c, the Co 2p spectra show two spin–orbit peaks (Co 2p3/2 at 779.5 eV and Co 2p1/2 at 794.5 eV) along with their shake-up satellite peaks. The Co 2p3/2 peak of Co100 mainly consists of Co3+ (779.5 eV) with a slight portion of Co2+ (780.5 eV).55,56 This Co2+ portion in the pristine sample seemingly arises53 from the formation of impurity phases due to the evaporation of Li during synthesis. After cycles with 4.6 V cut-off, the peak corresponding to Co2+ at Co 2p3/2 greatly increases and its satellite peak also shifts towards lower binding energy. These results are consistent with the formation of spinel phases such as Co3O4 and LixCo2O4. However, unlike Co100, the Co2+ and satellite peaks of Co97 and Co95 do not change significantly.
Fig. 8d–f show the high-resolution TEM images of Co100, Co97, and Co95 after the high voltage cycles, respectively. The yellow and purple boxes correspond to selected areas in the surface and bulk, respectively. The bulk FFT pattern of the cycled Co100 taken along the [110] zone axis exhibits a diffraction pattern indicative of a layered structure, while additional spots appear in its surface FFT pattern. These additional spots marked in yellow circles correspond to the (1−13) and (1−15) lattice planes of spinel phases. The spinel atomic arrangement is observed along the surface of each Co100 particle, as the phase borderline is drawn in white. By contrast, the spinel phase is less frequently observed around the surface of Co97 (Fig. 8e), and is not observed at all in the case of Co95 (Fig. 8f). These TEM results are consistent with those of the ΔS analysis in that Ni-doping inhibits structural degradation to spinel phases in high voltage cycles.
The coin cells were prepared in an Ar-filled glove box by assembling a Celgard 2400 separator, the LiCoO2-based electrode (working electrode), and Li metal foil (reference/counter electrode). 1 M lithium hexafluorophosphate (LiPF6) solution in a mixture of ethylene carbonate (EC) and dimethyl carbonate (DMC) (EC:DMC = 1:1, vol/vol) (PANAX E-TEC) was used as electrolyte. For battery testing, the galvanostatic mode was applied using a battery tester (PEBC05-0.01, PNE solution). The cell tests were conducted in the potential range of 3.0–4.6 V vs. Li/Li+ with constant current (CC) mode for both charge and discharge. For typical galvanostatic cycling tests, the coin cells were pre-cycled at 10 mA g−1 for 2 cycles, and the subsequent cycles were charged and discharged at 0.6C (1C = 170 mA g−1).
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c9ee02964h |
‡ H. J. Kim and Y. Park contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2020 |