Correlating Cation Mixing with Li Kinetics: Electrochemical and Li Diffusion Measurements on Li-Deficient LiNiO₂ and Li-Excess LiNi_0.5Mn_0.5O₂

Chemicals and materials

Synthesis of Li-deficient LNO

Synthesis of Li_1+x(Ni_0.5Mn_0.5)_1−xO₂

X-ray diffraction (XRD)

Scanning electron microscopy (SEM)

Coin cells

Atlung method for intercalant diffusion (AMID)

Figure 1 displays the SEM images of the synthesized powders with different Li/Ni ratios. All samples showed a typical spherical morphology of agglomerated primary particles with similar secondary particle sizes of ∼13 μm. Varying the Li/Ni ratio did not affect the particle size or morphology of the samples after the lithiation process. Higher resolution images, shown in Fig. 2 for selected samples, show that the average primary particle size was between 0.2−0.25 μm for all the samples measured. The secondary and primary particle sizes will be used later for measurements of Li chemical diffusion coefficients.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 3 shows the full XRD patterns and two expanded views for the entire series of synthesized materials. All samples were single phases with well-developed layered structures. No impurities were observed in the samples as seen from a lack of peaks in the lithium impurity region between 20–35 degrees. The peaks became sharper with clearer K_α1 and K_α2 doublets when increasing the Li/Ni ratio as can be seen from the expanded views of the (104) peak between 44^o–45^o and the (018)/(110) peaks between 64^o–65.5^o. This indicates a greater degree of crystallinity in the samples with higher Li/Ni ratio as the stoichiometric chemical composition is approached. Overall, this represents a set of highly crystalline materials with similar morphologies well-suited for comparisons of electrochemical and kinetic properties.

Zoom In Zoom Out Reset image size

Figure 4a shows the relationship between the expected and observed percentage of Ni atoms in the Li layers (%Ni_Li) for the various Li-deficient LNO powders. Expected values were calculated based on the amounts of Li and Ni used in the lithiation and charge balance. As the Li/Ni ratio is decreased during synthesis, the %Ni_Li is expected to increase linearly; the excess Ni is expected to assume a 2+ oxidation state, favorably occupying sites in the Li layers. The observed %Ni_Li values were close to the expected values except at high Li/Ni ratios, which is not surprising since there is always between ∼1%–2%Ni_Li even under Li-excess synthesis conditions (off-stoichiometry). To our knowledge, a pure two-dimensional structure with no Ni atoms in the Li layers has never been achieved with LNO. Similarly, in Fig. 3b, as the observed %Ni_Li increases, both the a and c lattice parameters increase due to the presence of Ni²⁺ ions in both Li and Ni layers; for each Ni²⁺ in the Li layer there must be one Ni²⁺ in the Ni layer to maintain charge neutrality. Since the Ni²⁺–O bond length is slightly longer compared to the Ni³⁺–O bond length, the average unit cell size is increased in all 3 directions. All the parameters used and obtained from Rietveld refinement are summarized in Table I.

Zoom In Zoom Out Reset image size

Figure 5 shows the 1st and 2nd cycle voltage vs specific capacity profiles for each Li-deficient LNO material compared to standard LNO (Li/Ni = 1.02) at C/20 between 3.0–4.3 V vs Li⁺/Li. The cells with greater Li/Ni ratios (smaller %Ni_Li) showed larger charge and discharge specific capacities and lower irreversible specific capacities (IRC) compared to the cells with smaller Li/Ni ratios (larger %Ni_Li). The clear plateaus and steps, representing phase transitions and single-phase regions, can be clearly seen in the higher Li/Ni ratio cells suggesting that the materials undergo Li insertion/extraction with faster Li diffusion, thus less kinetic hindrance. The impact of the %Ni_Li on the Li chemical diffusion, which is the main objective of this work, will be discussed in relation to Figs. 9–13.

Zoom In Zoom Out Reset image size

To understand the electrochemical reactions, the differential capacity vs voltage (dQ/dV vs V) during the 1st discharge and 2nd charge is shown in Fig. 6. The LNO cell showed typical phase transitions including hexagonal to monoclinic (H1 → M), monoclinic to hexagonal (M → H2), and hexagonal to hexagonal (H2 → H3), respectively. ¹⁹ The region below 3.6 V vs Li⁺/Li is often referred to as the kinetic hindrance (KH) region since the suppression of the dQ/dV peak has been shown to be caused by kinetic hindrance and can be recovered, in most cases, by cycling at higher temperature or lower rates. ^15,20 It is clearly seen that as more Ni atoms were added into the Li layers (smaller Li/Ni ratios), the kinetic hindrance peaks were gradually suppressed, suggesting that the presence of Ni_Li impacts the kinetic properties of the material. In general, all the peaks became less intense and broader, especially the H2 → H3 phase transition peak. This suggests that there was no abrupt layer shrinkage, typically observed in the presence of the H3 phase, possibly because of the presence of randomly substitutive Ni atoms in the Li layers that serve as pillars and prevents Li ordering, thus suppressing the H2 → H3 and other phase transitions. ^7,21

Zoom In Zoom Out Reset image size

Figure 7a shows the cycling performance of all samples at C/5 after the first two cycles at C/20. Standard LNO (Li/Ni 1.02) showed the largest specific discharge capacity as compared to the Li-deficient samples where the capacities decreased as the Li/Ni ratios decreased (%Ni_Li increased). The existence of extra Ni²⁺ in the Li layers requires the presence of^2x Ni²⁺ and (1−x) Ni³⁺ ions in the overall structure to maintain electroneutrality according to the Li_1−xNi_x ²⁺[Ni_1−x ³⁺Ni_x ²⁺]O₂ formula. This reduces the amount of electrochemically active Ni³⁺ available during charge/discharge. ²² There was little variation in capacity retention for all stoichiometries after 20 cycles despite the suppression of the H2–H3 phase transition in the samples with lower Li/Ni ratios, which is believed to be a major cause of poor capacity retention. ²³ Longer-term cycling and material characterization would provide useful insight into the capacity fade mechanisms in Li-deficient materials and more generally in materials with a larger %Ni_Li. Figure 7b shows the 1st cycle IRC at C/20 and the %Ni_Li as a function of Li/Ni ratio. The %Ni_Li and 1st cycle IRC follow very similar, near-linear trends, suggesting a strong correlation. Taken together with the capacity-rate dependence illustrated in Fig. 7a (the difference between C/20 and C/5 capacities), it is clear that the %Ni_Li strongly influences Li kinetics.

Zoom In Zoom Out Reset image size

Figures 8a and 8b show voltage vs time and voltage vs specific capacity from an AMID measurement. The cells were first charged and discharged at C/20 for 1 cycle followed by a C/40 charge. Sequential discharge currents as in the "signature curve method" ²⁴ ranging from 5C to C/160 with 1 h relaxation periods (OCV) in between were then applied in 0.1 V intervals from 4.3 to 3.6 V and in one voltage step for the KH region between 3.6–3.0 V. A larger voltage interval was chosen for the KH region because of the small amount of available capacity below 3.5 V, especially in the cells with smaller Li/Ni ratios. The cells were then charged back to 4.3 V at C/40. Figure 8b shows that there was no capacity slippage when the cell was charged back to 4.3 V after the AMID measurement. This confirms that there were no negative effects nor degradation of the active materials and that the capacities measured during the AMID cycle were real and not an artefact of side-reactions (i.e., electrolyte oxidation, capacitive effects, etc.). This indicates the importance of UHPC to accurately measure data over more than two orders of magnitude in C-rate to have confidence in the measurements.

Zoom In Zoom Out Reset image size

Figure 8c shows an example "signature curve" for the 3.6–3.0 V interval. "Signature curves" are a way of achieving specific states of charge corresponding to a given voltage in substantially less time compared to using one single very small current (to minimize kinetic effects). The cumulative capacity achieved when discharging to a specific voltage cut-off by sequentially lowering the current with intermittent open-circuit periods has been shown to be nearly identical to the capacity achieved using a single discharge with the smallest current used in the sequence. ²⁴ The protocol C-rates used for the AMID measurements were calculated based on the full theoretical capacity of the materials over the entire voltage range (3.0–4.3 V). However, since the AMID measurement involves discharging within specific voltage intervals each with different available capacities, effective rates are calculated as the total capacity achieved in a given voltage interval after the "signature curve" (sequence of discharge currents) divided by each respective current. Figure 8d shows the cumulative capacity extracted from each discharge current for the corresponding 3.6–3.0 V interval. The curve in Fig. 8d begins to plateau at low rates (far right) but is not exactly flat, suggesting that more capacity may have been extracted if even lower rates were used. This, of course, means that in cases where the capacity-rate curve does not fully plateau, the effective rates are not exact. In general, the capacity-rate curves showed a nice plateau at low rates, thus the effective rates are a good representation of the load on the cell.

To determine the voltage-dependent diffusion coefficient from the "signature curves," the capacity-rate data for each voltage interval is fit with a least-squares optimization to the solution derived by Atlung et al. ²⁵ to describe electrochemical insertion (or extraction) for spherical particles. Equation 1 is modified from Ref. 25 to include the discharge time and the possibility that the full available capacity may not have been extracted during a particular "signature curve" for a given voltage interval. Figure 8e shows the fitted experimental data (circles) on top of the theoretical curve developed by Atlung et al. on a dimensionless scale. The capacity-rate data from a "signature curve" for a given voltage interval, of the form (c in mAh, n in C/n), is fit to Eq. 1 treating both D_c and c_max as fitting parameters. The fractional capacity, c/c_max, is plotted against the dimensionless variable, Q = 3600 nD_c r⁻², which is the ratio of the discharge time, 3600 n, and the diffusion time, r²/D_c (r is the particle radius, thus r²/D_c is the characteristic time needed for a diffusing species to get from the surface to the center of a particle). A, B, and α_i are geometric factors, where for spherical particles A = 3, B = 5, and α_i cot (α_i ) = 1. The expansion included α_i up to i = 158 as this was found to be well-converged down to Q = 0.001. It is important to recognize that Eq. 1 is a solution to the phenomenological diffusion equations from Fick's Laws. Capacity-rate data extracted from a "signature curve" for a given voltage interval (it could be 4.3–4.2 V, or 4.0–3.6 V, for example) is numerically fit to Eq. 1 to extract D_c , which is an average quantity over the particular voltage interval. Examples of curve fittings and summary graphs showing quality of fit metrics for all samples are shown in Figs. S1–S9 (available online at stacks.iop.org/JES/168/090535/mmedia). Fine-grained details of the fitting procedure are given in Ref. 15.

$$\begin{eqnarray}&&{\rm{c}}/{{\rm{c}}}_{{\rm{\max }}}+\displaystyle \frac{{r}^{2}}{3600n{D}_{c}A}\left(\displaystyle \frac{1}{B}-2\displaystyle \sum _{i}\displaystyle \frac{\exp \left(-\left({\rm{c}}/{{\rm{c}}}_{{\rm{\max }}}\right)3600n{D}_{c}{\alpha }_{i}^{2}/{r}^{2}\right)}{{\alpha }_{i}^{2}}\right)=1\end{eqnarray} \tag{ 1 }$$

Figure 9 shows D_c as a function of discharge voltage for samples with Li/Ni ratios ranging from 0.88 to 1.02 obtained from the AMID measurements at temperatures of 20, 30, and 40 °C. To obtain the results in Fig. 9, a sphere radius, r, of 6.5 μm, characteristic of the secondary particle radius was used. Figure 10 shows D_c for the same experiments but where the primary particle radii, r, used in the fitting procedure were taken from the high-resolution SEM images shown in Fig. 2. A discussion is presented later to address whether using primary or secondary particle radii is most appropriate to use for this calculation. D_c for all samples increased as the operating temperature increased as is expected from classical diffusion theory. The samples with larger Li/Ni ratios (smaller %Ni_Li) showed larger D_c across the entire voltage range for all 3 temperatures with the only exceptions being the 3.0–3.6 V, and 3.6–3.7 V intervals at 20 °C. The dependence of D_c on the %Ni_Li is still clearly seen at 20 °C between 3.7–4.2 V. The data point for the 4.2–4.3 V interval is omitted in Fig. 9a (20 °C) because a large IR drop with very little cumulative capacity was measured in this interval, causing error in the calculation of effective rates and poor fits. These difficulties were not encountered for the 30 and 40 °C measurements (Figs. 9b, 9c, 10b and 10c).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

At all temperatures D_c was largest between 3.7–4.0 V when the interlayer spacing remained sufficiently large, maintaining the moderate Li tetrahedral site energy, while the Li concentration was sufficiently small to accommodate non-dilute di-vacancies. In the low voltage KH region (the 3.3 V data point), the %Ni_Li barely influences D_c . At such low voltages, the Li layers are nearly full and the di-vacancy concentration approaches zero, severely hindering Li diffusion. ²⁶ It is interesting that diffusion is already so slow below 3.6 V that the %Ni_Li has such little impact. D_c also becomes small beyond 4.2 V; the interlayer space is contracted as the Li layers are depleted, increasing the tetrahedral Li site energy due to coulombic repulsion from Ni atoms in the TM layer. ²

The data from Figs. 9 and 10 are re-plotted in Figs. 11 and 12 to show D_c as a function of %Ni_Li for voltages between 3.6–4.2 V and at 20, 30, 40 °C. Showing the data in this way, the impact of %Ni_Li on D_c is striking—in the 3.7–4.1 V interval, containing ∼80% of the total cell capacity, D_c decreases by an order of magnitude as the %Ni_Li goes from 2 to 6%. A 4% increase in the %Ni_Li leads to an order of magnitude decrease in D_c . In general, this trend is similar regardless of the temperature and regardless of the choice of secondary (Figs. 9 and 11) or primary (Figs. 10 and 12) particle radius. This fact alone explains the large increase in 1st cycle IRC observed in Fig. 7 as the %Ni_Li increased from 2%–6% and attributes the cause to poor Li kinetics. Figures 11a, 12a, 11f and 12f show a less pronounced decrease in D_c as the %Ni_Li increases. D_c below 3.6 V is not shown in Figs. 11 and 12 since the impact of the %Ni_Li is much less prominent due to Li kinetics already being so poor. These results demonstrate that the %Ni_Li can have a significant impact on the rate capability and reversible capacity of Ni-rich layered positive electrode materials, and perhaps more generally, any intercalation electrode with a non-negligible amount of cation mixing. As a result, one must keep the amount of Ni atoms in the Li layers as low as possible in order to maximize D_c for good rate capability and optimized electrochemical performance.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In the present work, whether the primary or secondary particle radius was chosen to fit the capacity-rate data obtained from the AMID measurements was inconsequential; the same qualitative trend was observed in both cases. However, since the primary particles were about 2 orders of magnitude smaller than the secondary particles (13 μm compared to 0.2 μm), the fitted magnitudes of D_c differed by nearly 4 orders of magnitude (D_c scales as 1/r²). Additionally, all samples studied in this work had very similar primary and secondary particle sizes. The materials with smaller Li/Ni ratios (larger %Ni_Li) had slightly smaller primary particle sizes compared to the materials with larger Li/Ni ratios (larger %Ni_Li)—0.2 μm compared to 0.25 μm. This magnified the differences in D_c slightly, as can be seen by comparing Fig. 9 with Figs. 10, and 11 with Fig. 12. In our previous work, the D_c for materials with the exact same composition but different particle morphologies, polycrystalline vs single crystalline, were compared. ¹⁵ The rate-capacity data obtained from both materials were nearly identical, indicating that r² /D_c should be the same. In order to obtain the same D_c for the polycrystalline and single crystalline materials, it was necessary to use the primary particle size when fitting the rate-capacity data from the polycrystalline material. This presents strong evidence that grain boundary diffusion is much faster than solid-state diffusion, thus the effective characteristic diffusion length is determined by the size of the primary particles. In this work, we chose to show D_c data obtained using both secondary and primary particle radii to convince the reader that this choice did not affect the dependence of D_c on the amount of Ni in the Li layer. However, when comparing D_c data obtained using the AMID for different chemical compositions and/or particle morphologies, it is crucial to be convinced that grain boundary diffusion has negligible contribution to Li kinetics and to use the primary particle radius in the fitting procedure. Below, a comparison of D_c in Li-excess ${{\rm{Li}}}_{1+{\rm{x}}}{\left({{\rm{Ni}}}_{0.5}{{\rm{Mn}}}_{0.5}\right)}_{1-{\rm{x}}}{{\rm{O}}}_{2}$ is presented.

To give a practical case study, Fig. 13 shows D_c as a function of discharge voltage and as a function of the %Ni_Li for ${{\rm{Li}}}_{1+{\rm{x}}}{\left({{\rm{Ni}}}_{0.5}{{\rm{Mn}}}_{0.5}\right)}_{1-{\rm{x}}}{{\rm{O}}}_{2}$ with x = 0, 0.04, 0.08, 0.12. This class of materials is of particular interest because it contains no Co and a large amount of Mn, thus has low material cost, and the specific capacity can be tuned by increasing the amount of excess Li (increasing x). Generally, Mn substitution in Ni-rich layered oxides without Li-excess promotes cation mixing because Mn is believed to preferentially accommodate a 4+ oxidation state, thus requiring a Ni²⁺ ion to balance the net charge, and divalent Ni ions, in the dilute regime, are believed to preferentially occupy Li sites. This simple argument is not exact; since as the Mn content becomes sufficiently large, the amount of cation mixing reaches a plateau. This makes sense since not all Ni atoms can occupy Li sites as this would leave semi-empty TM layers; the atomic ratios during synthesis encourage TM atoms to occupy the TM layers. However, as the excess Li content is increased in ${{\rm{Li}}}_{1+{\rm{x}}}{\left({{\rm{Ni}}}_{0.5}{{\rm{Mn}}}_{0.5}\right)}_{1-{\rm{x}}}{{\rm{O}}}_{2},$ there are at least x Li atoms occupying TM sites. For the net charge to balance, two Li atoms occupying TM sites promote one Ni atom from 2+ to 3+ (i.e. there are x/2 Ni³⁺). The increased concentration of Ni³⁺ lowers the %Ni_Li. This series of materials presents a perfect use-case to demonstrate more generally how the %Ni_Li impacts Li chemical diffusion in an industrially relevant positive electrode chemistry.

Zoom In Zoom Out Reset image size

Figure 13a shows D_c as a function of voltage measured at 30 °C from 3.6 − 4.4 V for each value of x. The primary particle size of each sample was used to determine D_c from the AMID measurements. D_c clearly increased with x; more excess Li led to larger D_c across the entire voltage range. Interestingly, the measured D_c for x = 0.08 and x = 0.12 were nearly identical suggesting that at high Li-excess there may be other diffusion mechanisms at play. It is important to keep in mind that as x increases, the amount of Li in the TM layer increases (even when x = 0 some Li atoms are in the TM layers to maintain electroneutrality)—it is possible that Li ions occupying TM sites provide inter-layer diffusion pathways that can circumvent Ni atoms in the Li layer. To draw analogy to the Li-deficient LNO materials presented in this work, Fig. 13b shows how D_c depends on the percentage of Ni atoms in the Li layers and x at various voltages. In this case, an increase in the %Ni_Li from ∼5% to ∼10% is accompanied by an order of magnitude decrease in D_c —the same trend that was observed in the Li-deficient LNO materials. However, when comparing the magnitudes of D_c from Figs. 11–13a, the LiNi_0.5Mn_0.5O₂ material, which has Li layers occupied with more than 10% Ni atoms, has a comparable D_c to the LNO-0.88 material, which contains about 6% Ni atoms in the Li layers. There are two considerations here: the first was discussed above—the presence of Li atoms in the TM layers may provide alternative diffusion pathways, the second is that the Li tetrahedral site energy may be decreased for tetrahedra near Mn atoms in the TM layers due to shorter Mn⁴⁺–O bond lengths, lowering the Li migration barrier. This work demonstrates that the presence of Ni atoms in the Li layers reduces D_c but that there are other mechanisms that can, at least partially, alleviate this hindrance. This is an important result that suggests practical limitations in primary particle size, in either single crystal or polycrystalline positive electrode materials, given rate capability requirements for any material where the %Ni_Li may be either sufficiently large after synthesis or may increase after prolonged cycling.

The impact of Ni in the Li layer (Ni_Li) on the electrochemical performance and Li chemical diffusion was demonstrated through a series of Li-deficient LNO materials where the amount of Ni_Li was carefully tuned. Increasing the amount of Ni_Li directly reduced the accessible specific capacity of the material and the 1st cycle IRC. Both effects were shown to be influenced by worsened Li kinetics. Measurements of the Li chemical diffusion coefficient using the recently developed "Atlung Method for Intercalant Diffusion" demonstrated a strong dependence on the amount of Ni_Li—a 4% increase in Ni_Li caused a ten times decrease in the Li chemical diffusion coefficient. This influence was most pronounced in the 3.6–4.2 V interval, which contains ∼80% of the cell capacity. These results systematically demonstrate the influence of transition metals in the lithium layer on the electrochemical and kinetic properties of a representative set of Ni-containing layered positive electrode materials. Considering a Li-excess series of materials (Li_1+x(Ni₅₀Mn₅₀)_1−xO₂) demonstrated generality; increasing the amount of Ni_Li led to smaller Li chemical diffusion coefficients, thus worse kinetics. These findings will provide useful guidance for both research and industry in the design of low-cost, high-energy positive electrode materials for LIBs.

The authors appreciate NSERC and Tesla Canada for the funding under the auspices of the Industrial Research Chair program. NP thanks the Vidyasirimedhi Institute of Science and Technology (VISTEC) for financial support. MC thanks NSERC for financial support. AL thanks the Walter C. Sumner Foundation for financial support. EZ thanks the Nova Scotia Graduate Scholarship (NSGS) for financial support.