Multiple diffusion pathways in LixNi0.77Co0.14Al0.09O2 (NCA) Li-ion battery cathodes

Department of Chemistry, University Colleg London, WC1H 0AJ, UK. E-mail: t.ashton@u ISIS Pulsed Neutron and Muon Source, Harwell Science and Innovation Campus, D Materials & Life Science Division, J-PARC C Frontier Research Center for Applied Ato Shirakata, Tokai, Ibaraki 319-1106, Japan Centre for Molecular and Materials Sc Vancouver, B.C., V6T 2A3, Canada Stewart Blusson Quantum Matter Institute, Mall, Vancouver, B.C., V6T 1Z4, Canada † Electronic supplementary informa 10.1039/d0ta03809a Cite this: J. Mater. Chem. A, 2020, 8, 11545


Introduction
As the global energy demand continues to rise, increased effort is directed towards the efficient harvesting (particularly from renewables), utilisation and storage of electrical energy to meet the needs of an increasingly urbanised global population and to alleviate the effects of climate change. 1,2 Towards the latter goal, the electrication of transport to move away from fossil fuels and linking this to smart charging grids will need development. Commercial rechargeable Lithium Ion Batteries (LIBs) with high energy densities are thus of interest in this regard, 3 most notably those with LiCoO 2 type layered metal oxide cathodes. 4,5 Such layered materials can be doped to improve performance characteristics, e.g. rate capability, operating voltage and structural/thermal stability.
Recently, layered, nickel rich LiNi 1Àx M x O 2 materials have been developed as inexpensive and more sustainable highenergy cathodes for secondary Li ion batteries (where x $ 0.2 and M is a substituent transition metal or combination of metals). Pure LiNiO 2 has shown poor performance as a cathode due to anti-site defects, where Ni 2+ species locate in Li 3a sites in the crystal structure and hamper Li + diffusion. 6 Improvement in performance was shown by substituting up to 20 at% Ni for Co, reducing the Li + /Ni 2+ cationic disorder and reducing oxygen loss at high states-of-charge and improving safety. 7 Further stabilisation strategies include the co-substitution of nickel with cobalt and another third element to form a ternary lithium containing metal oxide (LiNi x Co y M z O 2 , where x + y + z ¼ 1). Where M ¼ Mn, this is known as the widely researched "NMC" family of materials.
Specically, Al substituted materials such as LiNi 0.80 Co 0.15 Al 0.05 O 2 (NCA) are of increasing of interest for commercial battery applications due to its stable operating voltage of 3.8 V versus Li/Li + and observed gravimetric capacity between 175 and 200 mA h g À1 . 8 Inclusion of Al in the host structure leads to higher thermal stability and increased electrochemical properties. [9][10][11][12][13][14] Although the macroscopic electrochemical behaviour of NCA has been scrutinised in detail, the fundamental processes governing the operation of these materials is not well understood. Arguably, the most important of these processes remains the diffusion of lithium through the host insertion electrode, as lithium diffusion is ultimately responsible for cell operation. Many complementary techniques have been used to probe the diffusion mechanics; electrochemical (cyclic voltammetry, 15 titration techniques [16][17][18] and impedance spectroscopy 19 ), physical (nuclear magnetic resonance 20 and muon spin relaxation [21][22][23] ) and computational methods, 24-26 each with their own associated benets and limitations.
Muon spin relaxation (mSR) has previously been employed to investigate Li + diffusion in many LIB materials including cathodes and solid-state electrolytes. [21][22][23][27][28][29][30][31] As mSR is a local probe technique, atomic scale diffusion can be probed as a local average within the material. This allows the measurement of diffusion processes independent from contributions to diffusion encountered in bulk measurements, for instance intercalation-based processes arising from grain boundary and/ or surface interactions. The cathode materials that have been investigated to date include sub-stoichiometric Li x CoO 2 and LiNi1 3 Mn1 3

Co1
3 O 2 (NMC-111). 29,32 Values for the lithium diffusion rate (D Li ) determined by mSR in Li 0.73 CoO 2 and Li 0.53 CoO 2 of 7(AE2) Â 10 À10 and 2.5(AE0.8) Â 10 À10 cm 2 s À1 , respectively, were comparable to values calculated from rst principles. 24 However, these rst principles calculations suggested that decreasing lithium concentration could also promote a second Li + diffusion pathway (which has not been observed experimentally, to date). In these materials only one contribution to Li + diffusion was shown and assumed a single mechanism with an activation energy z 100 meV. 33 In comparison, previous bulk electrochemical studies using GITT on commercial NCA shows a D Li of 10 À10 cm s À1 and activation energies z 260 eV. 34 It is important to note that computational calculations and experimental methods have regularly overestimated E a of Li + diffusion when compared to local mSR measurements. 21,23 All of these studies have been based on a single contribution to Li + diffusion, however computational investigations on the parent material LiCoO 2 suggest two possible diffusion routes, namely the Oxygen Dumbbell Hop (ODH) and Tetrahedral Site Hop (TSH). 24 A representation of the two mechanisms can be seen in Fig. 1(a).
The ODH mechanism proceeds when a lithium migrates into an isolated vacancy (red circle), where the minimum energy path requires a lithium ion to pass between a dumbbell of two oxygen ions. Alternatively, the TSH mechanism occurs if the lithium ion can hop into a site that is part of a divacancy by passing through the tetrahedral site between the original position and the two vacant sites. The same mechanisms have also been calculated in silico to occur in Li x Ni 0.5 Mn 0.3 Co 0.2 O 2 , and suggest that metal site substitution of Co for other metals (such as Mn and Ni) can lead to a lowering of the activation energy of diffusion (120 meV) versus the analogous LiCoO 2 (210 meV). 35,36 The decrease in activation energy has been rationalised by two contributions: (i) weaker repulsive Li + -Ni 2+ interactions compared to Li + -Co 3+ ; (ii) a larger c lattice parameter. The relationship between the E a calculated for ODH and TSH consistently adhere to E TSH a z 1 4 Â E OSH a . Herein, the lithium diffusion mechanics in lithium decient NCA were investigated by bulk electrochemical impedance spectroscopy (EIS) and local mSR measurements. The mSR behaviour has provided the rst experimental evidence for the presence of two distinct lithium diffusion mechanisms in NCA cathode materials.

Structural characterisation
Sub-stoichiometric Li n Ni x Co y Al z O 2 particles were prepared by solid-state synthesis, described in the Experimental section. The syntheses yielded ne black powders which were characterised by Powder X-ray Diffraction (PXRD). The crystal structures of the NCA materials were analysed by subsequent Rietveld renement of the PXRD data shown in Fig. 2. A single rhombohedral phase, typical of LiCoO 2 -type layered materials (space group: R 3m), was observed in both cases and showed good agreement with the expected prole from an NCA reference pattern (ICSD no. 257247). 37 Lattice parameter values showed <0.5% difference compared to the reference structure The transition metal stoichiometry of both samples was found to be Li n Ni 0.77 Co 0.14 Al 0.09 O 2 using X-ray Fluorescence Spectroscopy (XRF). Li + /Ni 2+ cation mixing is a prevalent cause in disrupting Li + diffusion in layered transition metal oxides and was thus investigated by Neutron Powder Diffraction (NPD) and subsequent Rietveld renement, for both samples. Firstly, the ts were constrained to adhere to (Li gÀd Ni d ) 3b (Li d Ni 0.77 Co 0.14 Al 0.09 ) 3a O 2 ( Table 1); g is the occupancy of the atoms, g is the estimated Li + concentration from NPD and d is the measure of atoms participating in cation mixing and B is the Debye-Waller factor (attenuation of neutron radiation by thermally induced motion of the atoms). The metals composition was xed at the values obtained from XRF for each sample.
During the initial tting, the model could not converge with a positive value of d for either sample suggesting no detectable Li + /Ni 2+ cation mixing was present in either sample. Fixing d to 0 allowed the model to converge and a good t to lattice parameters taken from a rhombohedral R 3m (ICSD no. 257247) reference structure was obtained (Fig. 3) for both Li 0.87 NCA (R wp ¼ 2.66, R e ¼ 0.564, R F ¼ 2.58) and Li 0.71 NCA (R wp ¼ 3.09, R e ¼ 0.624, R F ¼ 2.79). The Li + concentration in each sample was estimated to be Li 0.87 NCA g ¼ 0.8704 (11) and Li 0.71 NCA g ¼ 0.7080 (12). A full table of calculated parameters can be found in the ESI. † The Li + concentration was xed at the determined values for each sample and further tting was carried out to estimate the lattice parameters, summarized in Table 2 with the PXRD values for comparison. The calculated lattice volumes marginally decreased by <0.1% with decreasing Li + content, in good agreement with previous crystallographic studies of NCA and NMC based materials. [38][39][40][41] No signicant difference in Li-Li distance was observed between Li 0.87 NCA (D ref ¼ 0.0055 A) and Li 0.71 NCA (D ref ¼ 0.0048 A) and was thus not expected to signicantly contribute to any difference in Li + diffusion kinetics.
The particle sizes of Li 0.87 NCA and Li 0.71 NCA were investigated further by scanning electron microscopy (SEM), seen in Fig. 4(a) and (b) respectively. The average particle size was calculated to be 3.4 mm (s ¼ 0.32 mm) for Li 0.87 NCA, while a bimodal distribution was observed for Li 0.71 NCA corresponding to average particle sizes of 2.4 mm (s ¼ 0.65 mm) and 0.73 mm (s ¼ 0.12 mm) for Li 0.71 NCA.

Bulk diffusion in Li x NCA (electrochemical characterisation)
Electrochemical impedance spectroscopy (EIS) was employed to investigate the bulk lithium diffusion in the lithium decient NCA samples. Fig. 5(a) shows Nyquist plots (real versus imaginary impedance) for Li 0.87 NCA and Li 0.71 NCA before and aer electrochemical cycling at a C/20 rate for 20 cycles. Aer cycling, both samples showed one semi-circle due to charge transfer resistance (R ct ) at the electrode surface, followed by a tail due to resistance from ionic transport (Warburg impedance). 42 Aer cycling, a second impedance contribution appeared due to the formation of a solid-electrolyte interphase (R SEI ) on the NCA particles, as described in previous reports on layered Li-ion cathodes. 43 R ct values of 51 and 59 U were extracted for Li 0.87 NCA and Li 0.71 NCA, respectively, by tting a modied Randles cell model (see ESI †) to the data before electrochemical cycling. 44 These R ct values were largely preserved aer cycling and the additional R SEI contributions were calculated as 51 and 54 U for Li 0.87 NCA and Li 0.71 NCA, respectively. Fig. 5(b) shows the Warburg plot for the two samples, where the gradient of the line is equal to the Warburg coefficient for the material (s w ). D Li for Li 0.87 NCA and Li 0.71 NCA were subsequently estimated using eqn (1), where: R is the gas constant, T is the temperature (298 K), A is the electrode area, F is the Faraday constant and C is the concentration of charge carriers.
Values of D Li were estimated as 3 Â 10 À11 cm 2 s À1 for Li 0.87 NCA and 1 Â 10 À11 cm 2 s À1 Li 0.71 NCA, in excellent agreement with previous EIS studies on NCA at similar Li + concentrations (D Li ¼ 4 Â 10 À11 for [Li + ] ¼ 0.75). 19 Local diffusion in Li x NCA (mSR) To investigate the diffusion mechanics on a local scale, mSR measurements were made from 150 to 400 K at 10 K intervals. At each temperature, three measurements were made at longitudinal magnetic eld strengths (B LF ) of 0 (zero-eld), 5 and 10 G to sequentially decouple the muon relaxation from nuclear magnetic elds within the samples (Fig. 6(a)). As B LF increases,

Atom
Site there is a decrease in the relaxation of the muon decay asymmetry suggesting that the decoupling was effective at the magnitude of applied elds chosen. At all values of B LF the slow relaxation of the data is indicative of the diffusion mediated perturbation of muon spin polarization from 6 Li and 7 Li. 21 Data from all three elds at a given temperature were t simultaneously to a modied Kubo-Toyabe relaxation model to describe the dynamic diffusion processes multiplied by a Gaussian relaxation function to account for temperature independent uctuations in Li + diffusion, and a constant background component. 45 Fig. 6 Table 1 based on standard data for NCA (space group R3m; ICSD coll. code: 257247). (c) Crystal structure of NCA showing the three main crystallographic sites of interest (000, 00z and 001/2) where Li atoms are green, MO 6 polyhedra are grey and O atoms are red. Table 2 Lattice parameters, cell volume and Li-Li distance calculated from PXRD and NPD. The neutron refinement data has also been given in full, with errors in the ESI, Table S1a and b NPD PXRD  for the local eld distribution width (D). A gentle decrease in D was observed due to the motional narrowing of the magnetic environment the muon experiences, similarly observed in other mSR diffusion experiments. Importantly no unexpected magnetic behaviour that would affect the calculated values of the muon relaxation uctuation rate (n) which can be seen in Fig. 7(a), were observed. Both Li 0.87 NCA and Li 0.71 NCA display two contributions to the uctuation rate in the low T range (160 to 260 K) and high T range (340 to 380 K). Although these observations cannot be explicitly assigned to their particular processes by mSR alone, two energy barriers have been predicted by the previously discussed theoretical work: a more rapid TSH mechanism and a less rapid ODH mechanism. 24 Therefore, the lower T contribution was assigned to D TSH Li and higher T to D ODH Li .

(b) shows the extracted values
As temperature increases the thermal equilibration and eventual homogeneity of Li + vacancies lead to a continual reduction in diffusion via TSH, due to the dilution of divacancies, and eventual dominance of the ODH pathway. Upon review of similar muon experiments on isostructural materials (e.g. LiCoO 2 , LiCrO 2 and LiNiO 2 ) a contribution at low T may also be present, although much weaker in magnitude. 28,33 Previous studies have found that the inclusion of Al into LiNi 0.5 Co 0.5 O 2 increases the Li + diffusion coefficient of the host material. 46 We therefore suggest this increase occurs by promoting the more facile TSH mechanism, leading to the exaggerated contribution observed by mSR versus other materials.
The lithium diffusion coefficients (D Li ) for both TSH and ODH can be calculated with eqn (2), where N i is the number of accessible Li + sites in the i th path, Z n,i is the vacancy fraction of the destination sites, s i is the jump distance between Li + sites, and n is the uctuation rate at each temperature. 47 For ODH, s i is equal to the a lattice parameter calculated from NPD (2.87 A) and for TSH it is equal to a Â O3 (4.97 A).
The D ODH Li and D TSH Li were calculated to be 3 Â 10 À11 and 3 Â 10 À10 cm 2 s À1 , respectively, for Li 0.87 NCA, and 3 Â 10 À11 and 2 Â 10 À10 cm 2 s À1 , respectively, for Li 0.71 NCA at 300 K. The values of D Li estimated from mSR agree with the values of D Li estimated using EIS and are further supported by rst principles calculations. Fig. 7(b) shows the Arrhenius plots for the muon hopping, giving rise to an E ODH a of 90 AE 13 and 65 AE 6 meV and E TSH a of 25 AE 1 and 21 AE 1 meV for Li 0.87 NCA and Li 0.71 NCA, respectively.  We observe that the activation energies for both hopping mechanisms are lower in Li 0.71 NCA, likely due to less Li-Li repulsion. The E a for TSH in both cases agrees well with values proposed for a divacancy mechanism in LiCoO 2 from PBE-SP calculations where values were as low as 15 meV. 48 The results obtained for the activation energies (E a ) for Li + diffusion for OSH and TSH were also close to previous computational studies, with respect to E TSH

Conclusions
For the rst time experimental evidence has been provided for the presence of two distinct diffusion pathways (ODH and TSH) operating in a layered intercalation cathode, NCA. Potential local structure contributions to multiple observed diffusion pathways (such as cation mixing and impurity phases) were ruled out using PXRD and NPD, and possible magnetic effects eliminated by scrutiny of D by mSR. Table 3 shows a summary of the values obtained from EIS and mSR. While both techniques are in good agreement with respect to estimated D Li , bulk techniques have been unable to distinguish between individual diffusion pathways highlighting the use of mSR as a powerful tool to understand technologically important/relevant materials. These results obtained from mSR are also in good agreement with values obtained from rst principles calculations on LiCoO 2 .
It has also been experimentally demonstrated that the TSH mechanism requires a lower E a and operates at a faster rate compared to the ODH mechanism. This is of much importance to energy materials research, as manipulation of layered cathode materials' structures to promote divacancies will lead to more facile diffusion, and ultimately more efficient battery operation. These experiments suggest that this may be achieved by cationic substitution with elements that promote the TSH mechanism, such as Al.

Preparation of lithium nickel cobalt aluminium oxide (NCA)
Samples of Li x Ni 0.8 Co 0.15 Al 0.05 O 2 were prepared via solid-state based methods. All raw materials were purchased from Sigma Aldrich (Dorset, UK) unless stated otherwise. Solid state synthesis was carried out by placing 3.15 g (99.9%) of Ni(NO 3 ) 2 $6H 2 O, 0.59 g (99.9%), of Co(NO 3 ) 3 $6H 2 O, 0.25 g (99.9%) of Al(NO 3 ) 3 $9H 2 O and 0.97 g (99.9%) LiOH in to a glass beaker. 30 mL of isopropanol was added to the mixture and the resulting brown slurry was mixed using a high shear mixer (IKA Ultra-Turrax T25) for 5 min at 10 000 rpm to give a homogenous mixture. The mixture was transferred to a ceramic crucible and placed in a box furnace for 30 min at 75 C to evaporate the isopropanol. The temperature was then ramped at 10 C min À1 to 850 C and held for 5 h, yielding a black solid material. Aer heat treatment, potential LiOH or Li 2 O impurities were removed by washing with 0.05 M acetic acid (1 Â 30 mL), water (3 Â 30 mL) and isopropanol (1 Â 30 mL) by centrifugation and dried in a vacuum oven overnight at 60 C giving rise to sample Li 0.87 NCA. Li 0.71 NCA was produced by using an identical methodology using 4 g of nanoparticle precursor (see ESI †) in place of the nitrate precursors.

Physical characterisation
Powder X-ray Diffraction (PXRD) patterns were collected using a STOE Stadi P diffractometer (Mo Ka 1 radiation, l ¼ 0.70932 A) in transmission geometry equipped with a germanium (111) monochromator and a DECTRIS Mythen 1K silicon strip detector (Dectris, Baden, Switzerland). An yttria (Y 2 O 3 ) standard was used to estimate instrumental peak broadening. Datasets were collected over the 2q range of 2 to 40 with a step size of 0.5 and a count time of 20 s per step. Rietveld renement was performed using the FullProf™ soware package. 49 Neutron powder diffraction (NPD) was carried out on the iMATERIA diffractometer at J-PARC in Tokai, Japan. 50 The data was analysed by Rietveld renement using the Z-Rietveld soware package. 51,52 Scanning Electron Microscopy (SEM) was performed using a JEOL JSM-6700F microscope. To minimise charging samples were deposited on copper foil tape, mounted on aluminium stubs, aer dispersion in methanol (99.9%, Sigma Aldrich, Dorset, UK) and ultrasonication in a XUBA3 Ultrasonic Bath (Grant Instruments, Cambridge, UK) for 5 minutes. Image analysis was carried out using ImageJ soware, calculating the average size over 200 particles for each sample.
X-ray uorescence (XRF) spectroscopy was performed on the synthesised powders using an Epsilon 4 spectrometer (Malvern Panalytical, Malvern, UK). A nominal <2% total relative error applies to the observed values.

Electrochemical characterisation
Electrodes were prepared by mixing of the active material (Li x NCA) with binder (polyvinylidene uoride, PVDF, PI-KEM, Staffordshire, UK) and conductive carbon black (Super P, Alfa Aesar, Heysham, UK) in a ratio of 80 : 10 : 10 wt%. A 10 wt% solution of PVDF in N-methyl-2-pyrrolidone, NMP (Sigma Aldrich, Dorset, UK) was prepared using a magnetic stirrer and then mixed by hand with the conductive carbon and the active material. Further NMP (2.5 mL) was added to give a viscous slurry. The slurry was ball-milled at 800 rpm for 1 h before being cast on 15 mm thick aluminium foil (PI-KEM, Staffordshire, UK). The electrode sheets were dried on a hotplate at ca. 150 C for 20 min.
All half-cell testing of the NCA materials were performed in CR2032 coin cells at room temperature on a Gamry Instruments Interface 1000 Galvanostat/Potentiostat (SciMed, Cheshire, UK). All cells were assembled in an Ar-lled glovebox with O 2 and H 2 O < 0.5 ppm. GF/B glass microber lters (Whatman, Buckinghamshire, UK) were used as separators and drenched in electrolyte of 1 M LiPF 6 in 1 : 1 wt% ethylene carbonate/diethyl methyl carbonate (BASF, Ludwigshafen, Germany). Lithium metal foil 15.6 mm diameter, 0.45 mm thick (PI-KEM, Staffordshire, UK) was used as a counter electrode. Potentiostatic EIS was carried out from 1 to 10 6 Hz at 10 mV AC RMS and 0 V DC versus open circuit voltage.

Muon spin relaxation (mSR)
Muon spin relaxation experiments were performed at STFC ISIS Neutron and Muon Source, Didcot, UK. Approximately 3 g of powdered samples of Li 0.87 NCA and Li 0.71 NCA were packed into recessed titanium samples holders, covered with a titanium window and secured with a titanium bezel. The titanium holder provides a simple background signal from muons which are not implanted into the sample and is easily subtracted from the data. The sample holder was placed into the muon spectrometer (EMU) and evacuated to <1 Â 10 À6 mbar. The instrument was cooled to a temperature of 160 K and the instrument asymmetry was measured using a transverse magnetic eld of 20 G. Measurements were taken every 10 K up to 380 K with longitudinal at elds of 0, 5 and 10 G. Data collected at each temperature for the three applied magnetic elds were t simultaneously using the WiMDA soware. 53

Conflicts of interest
There are no conicts to declare.

Note added after first publication
This article replaces the version published on 28th May 2020, which contained errors in the Fig. 4 and Fig. 5 captions.