Direct Observation of Dynamic Lithium Diffusion Behavior in Nickel-Rich, LiNi0.8Mn0.1Co0.1O2 (NMC811) Cathodes Using Operando Muon Spectroscopy

Ni-rich layered oxide cathode materials such as LiNi0.8Mn0.1Co0.1O2 (NMC811) are widely tipped as the next-generation cathodes for lithium-ion batteries. The NMC class offers high capacities but suffers an irreversible first cycle capacity loss, a result of slow Li+ diffusion kinetics at a low state of charge. Understanding the origin of these kinetic hindrances to Li+ mobility inside the cathode is vital to negate the first cycle capacity loss in future materials design. Here, we report on the development of operando muon spectroscopy (μSR) to probe the Å-length scale Li+ ion diffusion in NMC811 during its first cycle and how this can be compared to electrochemical impedance spectroscopy (EIS) and the galvanostatic intermittent titration technique (GITT). Volume-averaged muon implantation enables measurements that are largely unaffected by interface/surface effects, thus providing a specific characterization of the fundamental bulk properties to complement surface-dominated electrochemical methods. First cycle measurements show that the bulk Li+ mobility is less affected than the surface Li+ mobility at full depth of discharge, indicating that sluggish surface diffusion is the likely cause of first cycle irreversible capacity loss. Additionally, we demonstrate that trends in the nuclear field distribution width of the implanted muons during cycling correlate with those observed in differential capacity, suggesting the sensitivity of this μSR parameter to structural changes during cycling.


S-2
Rietveld refinement of as-synthesized NMC811 Figure S1. Rietveld refinement of as-synthesized LiNi0.8Mn0.1Co0.1O2 to a R3 ̅ m layered oxide crystal structure after calcination at 850 ˚C for 12 h in an O2 atmosphere. A weighted residual of 1.81% was found.

BAM (Battery Analysis by Muon) operando cell assembly
To achieve a reasonable signal during the operando experiment, a significant mass of active material (≥100 mg) was required to ensure adequate muon implantation. The carbon content of the prepared cathode powders (70:20:10, NMC811:C:PTFE wt.%) was high to ensure sufficient electronic conduction across the large area/mass. The devised methodology for this experiment involved the usage of ~50 mg of cathode powder in layers, packed down using a spatula. Small glass-microfibre separator semicircles were used alongside separator rings to hold additional electrolyte within the cell to fully wet all cathode particles and enable high-rate cycling, although large contact areas between cathode layers were retained. This cell assembly procedure was found to display comparable performances to standard low mass loading coin/Swagelok type cell configurations over the first few cycles (i.e., ~200 mAh g -1 first discharge capacity) and produced consistent cycling results. S-4 Figure S2 displays a comparison of different cell types and cathode forms. As can be clearly seen, the first cycle is very comparable between the coin, Swagelok, and BAM cells, highlighting the reproducibility of electrochemistry in our developed operando cell.
Normalised dQ/dV plots also show very similar peak positions. As the active mass loading is increased to a level required for operando μSR measurements (~70 mg cm -2 ) there is a small overpotential noticeable in the cycling and dQ/dV profile. This is expected given the unavoidable increase of internal resistance as a consequence of the greater mass loading. As such, caution is expelled when comparing datasets from different measurement techniques on different cells, as the voltage of structural and electron transfer events during cycling will be slightly shifted.

S-5
Muon Spin Relaxation Measurements Figure S3. Stopping profile of muons in the BAM cell for the operando experiment, as simulated by the SRIM program. The muon is approximated as a light proton for this simulation. The path of the muon in the cell can be visualised in Figure 2d, where the stainless steel current collector is directly before the cathode. For this simulation, the cathode is assumed to be homogeneously interspersed with the electrolyte. Although the exact thickness of the cathode is not known because the cell is under compression, the mass loading is correct, which is the important factor when considering the stopping behaviour of the muons. The simulation shows that all muons are predicted to stop within the cathode layer, implanting within all its constituents. Very few muons are likely to reach the separator. We can thus be confident that most muons will stop in the cathode/electrolyte layer within the cell.

Operando µSR Fitting
Muon measurements were fitted using Eq. 1. The exponential term was added to account for the other materials contained within the operando cell (carbon, binder, etc.), which some muons will undoubtedly stop within. Due to the low volume fraction of active material, the flat background was found to be a large value of 0.159 (2). A relatively low Kubo-Toyabe amplitude of 0.034 (2) was obtained (these are displayed in Table S3), which is around 17% of the total measured amplitude. The fitting function was constrained to three separate runs,  (Table S4) but seem to slightly deviate at extreme potentials (i.e. high and low voltage). This is indicative of the structural changes within the sample and/or large voltage range during the measurements taken across these points. Such changes cannot realistically be modelled without a detailed ex situ experiment on isolated samples with specific compositions: something that will be worthwhile in future work. Thus, the model used in this work, with fixed component amplitudes across the cycle, represents the most reliable method to fit this operando data. Fitted asymmetry μSR data across the first cycle of NMC811 Li + mobility appears markedly slower at low SOC.

S-12
Operando XRD experiment Figure S6. The first charge of the operando XRD cell from the results shown in (a). An initial overpotential is seen, likely due to electrode surface impurities from potential air exposure and is not predicted to significantly affect any obtained XRD patterns. The EL-cell was cycled at C/50 and achieved a charge capacity of 227 mAh g -1 .
S-13   The change in resistance of the charge transfer (RCT) and cathode-electrolyte interface (RCEI) components as obtained by equivalent circuit fitting ( Figure S9).
S-16 Figure S9. Example of the fitted impedance spectra at 4.4 V for the data shown in Figure S8.
An [Rsol + RCEI/CPE1 + RCT/CPE2 + W1] equivalent circuit was used, as is common for impedance fitting of NMC half cells. 3 Rsol is defined as the solution resistance (between 0 Ω and first semi-circle), RCEI as the cathode/electrolyte interface resistance, and RCT as the charge transfer resistance through NMC secondary particles.
S-17 Table S4. Fitted values of resistance and capacitance for the equivalent circuit displayed in Figure S9 across the first cycle of an Li/NMC811 half-cell. These data are presented graphically in Figure 6d. The smaller capacitance of C1 (~10 −6 F) suggests this arises from phenomena at the sample-electrode interface or a passivation layer on the particle surfaces,

Galvanostatic Intermittent Titration Technique (GITT)
The chemical diffusion coefficient, Ds, was evaluated from GITT data by deriving a surface area independent term from the simplified Weppner-Hubbins experesion defined as: 5 Eq. S1 In Eq. S1, S is the active material surface area, τ is the time of the transient current pulse, nm is the number of moles of the active material, and vm is the molar volume of the active material.
These are assumed to be constant throughout the experiment. ∆Et and ∆Es are defined above as originating from the GITT measurement itself. ∆Et is the difference in voltage during the current pulse period, while ignoring the IR drop component (vertical rise in voltage). ∆Es is the difference in voltage at the end of consecutive relaxation steps. For this method to be valid, we ensure that τ << R 2 /Ds, where R is the diffusion length, which is in this case considered to be approximately equal to the average radius of a the secondary particles.

Dipolar Field Calculations and Muon Site Discussion
To better understand the muon stopping site in NMC811, dipolar field calculations were performed to compare theoretical with experimental Δ values. These were completed in a similar manner to the method previously used by our group, 6 using Eq. S2. 7 The unit cell size of z ≤ 0.5 from our previous study was increased to z ≤ 1.0 for these calculations. In Eq. S2, γμ is the muon's gyromagnetic ratio, Ii and γi are the spin and the nuclear gyromagnetic ratio of the i th nucleus, respectively, and ri is the distance between the i th nucleus and the muon site. This was calculated over the full unit cell for NMC811. The muon is predicted as roughly 1 Å away from any oxygen atom in the structure.