The Effect of Interdiffusion on the Properties of Lithium-Rich Core-Shell Cathodes

To obtain the highest energy density, lithium ion batteries based on layered Li-Ni-Mn-Co oxides need to be charged to above 4.5 V without sacriﬁcing lifetime. Lithium-rich core-shell (CS) structured positive electrodes having a high energy density core material with poor stability against the electrolyte can be protected by a thin layer of a stable shell material. In this work, the effects of the initial shell thickness, sintering temperature and interdiffusion during sintering on the electrochemical performance of CS cathodes were studied for the ﬁrst time to our knowledge. Full cell coin cells of selected CS samples with electrolyte additives showed good capacity retention. This work will help guide the design of next generation positive electrode materials using CS and coating strategies.©TheAuthor(s) 2016

Lithium ion batteries (LIBs) with high energy density, low cost as well as long-life time are required for large scale adoption of electric vehicles. The layered Li-Ni-Mn-Co oxides (NMC) are excellent positive electrodes candidates for cost effective LIBs. 1,2 A large operating voltage window (charge to 4.5 V and above in full cells against graphite) is needed to increase the energy density and minimal electrolyte oxidation is required so the life-time of the cells will not be sacrificed. 3 Besides the development of novel electrolyte systems using additives or new solvents, [3][4][5][6][7][8][9][10][11] coatings on the positive electrode material can minimize electrolyte oxidation in high voltage cells. [12][13][14][15][16] However, thick coatings (higher than 3 wt%) with inactive oxides on NMC materials significantly decrease the energy density and rate capability of the cells and thus these coatings are usually incomplete with nano-particles scattered on the NMC surface. 12,13 Core-shell (CS) structured positive electrode materials based on NMC could be the next generation of positive electrode materials for high energy density lithium-ion batteries. This is because a high energy core material with poor stability against the electrolyte can be protected by a thin layer of a stable and active shell material with lower Ni and higher Mn content. 17 Core-shell or gradient LiNi x Mn y Co z O 2 (x + y + z = 1) materials for voltages lower than 4.4 V were first developed by Y. K. Sun's group. [18][19][20] These have a high Ni content in the core and increasing Mn content from the core to the surface, with a maximum Mn content on the surface of ∼50%. In our previous report, 17,21 Li-rich and Mn-rich materials 22,23 were used as the protecting shell for voltages above 4.5 V. It was shown that the Li-rich and Mn-rich shell protected the Ni-rich core from reactions with the electrolyte while the Ni-rich core rendered a high and stable average voltage. 21 Diffusion of the cations between the core and shell phases occurs during sintering to prepare the NMC core-shell oxide. 24 Although much research on CS and gradient NMC materials have been reported, there are no detailed studies of how interdiffusion during sintering affects the final composition, structure and electrochemical properties of CS materials containing Li, Ni, Mn and Co. Previously, stacked laminar pellets were designed to study interdiffusion in a binary system (only two of the three elements, Ni, Co and Mn were used). 24 In this work, both the core and shell contained all of Li, Ni, Mn and Co so that complicated interdiffusion occurred. Concentration profiles measured on spherical CS particles and pre-designed laminar pellets were fitted with modeling results, improving the understanding of NMC CS materials and providing guidance for future improvements. reactor (CSTR) (Brunswick Scientific/Eppendorf BioFlo 310) using a method similar to that described by Van Bommel et al. 32 Aqueous solutions of NiSO 4 , MnSO 4 and CoSO 4 were prepared with Ni:Mn molar ratios of 2:1 for the core, and Ni:Mn:Co molar ratios of 1:3:1 for shell. The total metal ion concentration of each solution was 2.0 M. A 10.0 M NaOH (aq) solution was used as the source of base for the reaction, while a 5.0 M NH 3(aq) solution was used for metal ion coordination with ammonia to facilitate spherical and dense particle growth during the reaction. 32 A precursor with a Ni:Mn:Co ratio of 4:6:0 was also prepared for the pellet study.
Reagents were added using digital peristaltic pumps (Watson-Marlow IP31). Sodium hydroxide addition was automatically controlled by the pH controller and added as required by a peristaltic pump on the reactor. The vessel was maintained at a temperature of 60 • C and the contents of the reactor were stirred by an overhead stirrer at 600-800 rpm. Nitrogen was bubbled (60 sccm (standard cubic centimeter per minute) into the reactor throughout the reaction to create an inert reaction atmosphere in order to minimize the oxidation of Mn-rich hydroxides. The pH electrode (Mettler-Toledo InLab 424) was calibrated at room temperature using buffer solutions. The pH values of the buffer solutions were 4.0 and 10.0 at 20 • C (Fisher Scientific).
For core only and shell only precursors, a volume of 1 L of a 1 M NH 3(aq) solution was heated to 60 • C. The reaction proceeded with the addition of 5.0 M NH 3(aq) at 0.14 ml/min and 2.0 M MSO 4 at 0.5 ml/min. The total volume in the reaction vessel was allowed to gradually increase as the reaction continued, with a final volume of approximately 2 L. After 20 hours of reaction time, the suspension was cooled to room temperature and then filtered and rinsed with 4.0 L of water, followed by heating at 120 • C overnight in a forced air oven.
For the CS precursors, the core solution was pumped for ∼14 hours (400 mL) and then pumping switched to the shell solution for ∼7 hours (200 mL), which yields about 33 mol% shell (molar ratio). For 10 mol% and 20 mol% shell, the relative volume and times of the core and shell metal sulfate solution were adjusted accordingly. The desired pH values were changed accordingly when the solutions were switched from core to shell.
CS precursors with (Ni 0.6 Mn 0.2 Co 0.2 )(OH) 2 as the core and 10 mol% (CS10), 20 mol% (CS20) or 33 mol% (CS33) (Ni 0.2 Mn 0.6 Co 0.2 )(OH) 2 shell were first synthesized and dried. Each dried precursor was recovered, ground and mixed with a stoichiometric equivalent of Li 2 CO 3 (Umicore) by mechanical grinding until a homogenous consistency was achieved, about 10-15 minutes. Three grams of precursor was used for the synthesis of each sample. Lithiated samples were then prepared by sintering the precursor and LiOH with three different lithium contents (average Li with x= 0.02, 0.04 and 0.06 in Li 1+x TM 1-x O 2 22 ) at 850 or 900 • C for 10 h. The amount of LiOH was calculated based on the average composition and target lithium content. The samples were labeled as CS10 (20, 33) -850 (900) -1 (2,3), which indicate the initial shell content, sintering temperature and lithium content, respectively. For example, CS20-900-3 indicates a sample with 20 mol% shell, sintered at 900 • C with x = 0.06. The final products were mechanically ground and passed through a 75 micron sieve prior to characterization.
Characterization.-Elemental analysis (EA) used inductively coupled plasma optical emission spectrometry (ICP-OES) to determine the Mn, Ni and Co ratios in the precursors. Approximately 10 mg of each sample was dissolved in 2 mL 2:1 reagent grade HCl:HNO 3 aqua regia solution which was then diluted prior to measurement.
Powder X-ray diffraction (XRD) was carried out using a Siemens D5000 diffractometer equipped with a Cu target X-ray tube and a diffracted beam monochromator. Diffraction patterns were collected in the scattering angle (2θ) range of 15-70 • at 0.05 • intervals with a dwell time of 3 s. The lattice constants of the samples were shown in the supporting information.
A NanoScience Phenom Pro G2 Desktop Scanning Electron Microscope (SEM) was used to study the morphology of precursors and sintered samples. Samples were prepared by mounting the powder on adhesive carbon tape prior to imaging.
EDS mapping measurements for CS33-900-2 were prepared by first encasing powder in epoxy (CrystalBond 555, SPI Supplies/Structure Probe Inc.). The particles encased in epoxy were cut with sandpaper and then polished to a mirror finish with alumina paste. The stubs were then coated with amorphous carbon (∼40 nm thick) using magnetron sputtering. The mapping was carried out using a Hitachi S-4700 SEM equipped with an Oxford Instruments 80 mm 2 silicon drifted detector. Elemental maps of samples were collected in 300 seconds with an accelerating voltage of 20 kV and current of 15 μA.
Samples for EDS mapping measurements of CS33-850-2 were prepared by an ion beam cross-section polisher (JEOL IB-09010CP) at the Canadian Centre for Electron Microscopy (CCEM). The CS sample was first embedded in a graphite block with carbon paint (PELCO) and then the cross-section of the block was milled by the cross-section polisher with an Ar ion beam. The EDS mapping was carried out using a JEOL JSM-7000F SEM at CCEM. Elemental maps of the samples were collected in 300 seconds with an accelerating voltage of 10 kV and current of 10 μA.
EDS point analysis on CS-850-2 was carried out using an aberration-corrected (image and probe-forming lenses) FEI Titan Cubed 80-300 microscope operated in STEM mode with an acceleration voltage of 200 keV. The TEM sample was prepared using a dual beam focused ion beam/scanning electron microscope (FIB/SEM) (Zeiss NVision 40) to obtain a thin slice from the center of a particle, which was then mounted onto a FIB lift-out grid (PELCO) and eventually thinned down to ∼70 nm prior for the analysis. Figures S1a and b show electron trajectories for thin film LiNiO 2 (1000 nm) and the corresponding radial X-ray distribution for Ni-Kα, respectively, simulated for a 200 kV accelerating voltage and a 1 nm beam diameter using CASINO software. Figure S1b shows that the X-rays are generated primarily within an 80 nm radius around the center of the electron beam. were synthesized using the coprecipitation method and the CSTR for the precursors. The powders were mixed with 2 wt% of LiOH•H 2 O as a "binder" prior to making the pellets and to compensate for possible lithium loss during heating. For each pellet, powder A was first compressed to 15 MPa for 5 minutes, then powder B was added to the pellet mould on top of the pellet A. The powders were compressed again to 120 MPa for 10 minutes to ensure good contact at the interface. The pellets were heated to 900 • C for 10 h and then cooled slowly. The heated pellets were purposely fractured into smaller pieces, which were subsequently embedded into Crystal Bond (CrystalBond 555, SPI Supplies/Structure Probe Inc.) and polished to a mirror surface finish for scanning electron microscopy (SEM) and EDS experiments. The SEM studies were made on the cross sections of the diffusion couple pellets.
SEM evaluation was conducted using a Hitachi S-4700 SEM with a cold field emission source. The beam size is less than 10 nm. EDS line scans were collected with an accelerating voltage of 15 kV, a beam current of 15 μA, and an acquisition time of 20 seconds for each data point. The penetration depth of the electron beam was ∼800 nm, while the X-rays originate primarily from a radial distance of less than 80 nm from the electron beam. Between 20 to 50 points were collected for each diffusion couple, separated by a distance of ∼0.5 to ∼2.0 μm, depending on the sample. Two to four different line scans were measured for each sample. Measured concentration profiles from the same pellets were aligned and an average concentration profile was then calculated for least squares fitting to calculations made with Fick's law.
Electrochemistry.-Electrochemical measurements (half cells) were carried out via galvanostatic charge-discharge cycling using standard 2325 coin cells with lithium metal negative electrodes on an E-One Moli Energy Canada battery testing system. During electrode preparation, the slurry was a mixture of 92 wt% active material, 4 wt% Super-S carbon black (Timcal) and 4 wt% PVDF (Polyvinylidene fluoride, Sigma-Aldrich) with NMP (1-Methyl-2-pyrrolidinone, 99.5%, Sigma-Aldrich) as the solvent. Electrodes were prepared by coating the slurry on an Al foil with a 150 μm notch bar spreader. The electrodes were dried overnight at 120 • C in a vacuum oven before use. The electrolyte was 1.0 M LiPF 6 in 1:2 v/v ethylene carbonate:diethyl carbonate (EC:DEC) (BASF, max < 20 ppm water). The separators used were one Celgard 2320 (Celgard) on the lithium electrode and one polypropylene blown-microfiber separator (3 M) adjacent to the positive electrode. All the cells were tested with a specific current of 10 mA/g at 30 • C. The cells were tested between 2.5-4.4 V for 4 cycles, then 2.5-4.6 V for 10 cycles, followed by 2.5-4.8V for 1 cycle, and 2.5-4.6 V for another 10 cycles.
The full coin cells followed the same procedure except a graphite electrode from Magna E-car was used as the counter electrode. The control electrolyte was 1.0 M LiPF 6 in 1:2 v/v ethylene carbonate:diethyl carbonate (EC:DEC) (BASF, max < 20 ppm water). Electrolyte additives used in this work were a combinations of 2 wt% PES + 1 wt% MMDS + 1 wt% TTSPi (PES211). The cells were tested between 2.8-4.5 V for 100 cycles.
Ternary interdiffusion model.-In a ternary system, the flux of one component can be expressed in terms of two independent fluxes, which means the concentration change of one component with time is dependent on the concentration gradient of the element itself and another independent element. Fick's law can be expressed as: [33][34][35] Hence, there are four independent interdiffusion coefficients, and two independent concentration profiles. The superscript "3" in the notation of the interdiffusion coefficient indicates that the 3 rd element is dependent on the other two. The equation can be discretized by equidistant segments of size x and t, and solved numerically using the same method introduced in Ref. 24. Basically, Symmetric boundary conditions were assumed with the introduction of hypothetical points x −1 and x n+1 at both ends of the sample, which were set to have the same values as the points at positions x 2 and x n-1 , respectively. Concentrations at the next time step can be determined from the initial concentrations with Euler forward integration. ranges is defined as below: 33 Results and Discussion Figure 1 shows the XRD patterns of the as-synthesized hydroxide precursors of the core-only, shell-only, and CS samples with 10%, 20% and 33% shell, respectively. Figure 1 allows one to infer that the core and shell phases in the CS samples have the correct composition and structure. Table I shows the elemental analysis results which match very well to the average target composition. Figure 2 shows scanning electron microscopy (SEM) images and particle size analysis of the CS precursors. The average particle size is around 14-16 μm in the CS samples. Figure S2 shows the SEM images and energy dispersive spectroscopy (EDS) mapping results of CS precursors. Figure S2 shows a clear Mn-rich shell in CS10, CS20 and CS33. Figure 3 shows the SEM images and EDS mapping results of lithiated CS samples prepared at 850 and 900 • C for 10 h. Figures 3a1,  3b1 and 3c1 show the elemental mapping results for CS10, CS20 and CS33 samples with x = 0.04, respectively, sintered at 850 • C. Figures  3a1-3c1 show a clear Mn-rich shell in CS20-850 and CS33-850 while no obvious Mn-rich shell was observed in CS10-850. Additionally, the Ni and Co content at the surface is lower than that in the core (lower brightness in the EDS maps) in CS33-850. Figures 3a2, 3b2 and 3c2 show the mapping results for the corresponding samples prepared  SEM images of C (a1), CS10 (b1), CS20 (c1) and CS33 (d1) precursors and particle size analysis of C (a1), CS10 (b1), CS20 (c1) and CS33 (d1). The shell thickness was calculated by assuming spherical particles as introduced in Ref. 24. 900 • C. There was no obvious Mn-rich shell maintained in CS20-900 as there is in CS20-850. A clear Mn-rich shell was detected in CS33-900, indicating that sintering temperature and shell thickness have a significant effect on the final composition profile of the lithium-rich CS samples due to interdiffusion between the core and shell. In order to examine the interdiffusion phenomena in spherical CS particles with improved spatial resolution, a focused ion-beam (FIB) was used to cut a thin slice (∼100 nm) through the center of a randomly selected particle of CS33-850-2 for studies using EDS in scanning transmission electron microscopy (STEM). Figure 4a shows a STEM image of the prepared slice. The yellow line shows the path where EDS point analyses were performed. Figure 4b shows the measured concentration profiles with symbols, calculated profiles with solid lines and simulated initial concentration profiles with dashed lines respectively. The model based on Fick's law used for the fitting that describes the ternary diffusion phenomena is discussed in the Experimental section. Figure 4b shows that Ni moved from the core to the shell following the concentration gradient and that the Ni content on the surface changed from ∼21% to ∼30% during sintering, while Mn moved from the shell into the bulk, and the Mn content on the surface changed from ∼57% to ∼55%. Surprisingly, the Co content near the interface on the core side is higher than its initial concentration while it is lower on the shell side. This indicates that Co moved into the core from the shell, even though the initial Co contents in the core and shell were the same, in order to compensate for the increase of Ni content at the surface. This is because the interdiffusion between Ni/Co is much faster than Ni/Mn and Co/Mn as discussed in Ref. 24. This suggests that the presence of Co in the shell can accelerate the diffusion of Ni from the core to the shell.
and studied with EDS. A concentration profile across the interface can then be measured on the polished cross-section of the pellet surface using SEM/EDS point scans. Figure 4c shows the measured concentration profile of the NMC622 and NMC262 couple, while Figure S3 shows the SEM image and EDS mapping results near the interface. Figure 4c shows the same trend as discussed in the spherical case: Co moved from the shell to the core even there is almost no initial Co  concentration gradient. Additionally, Figure 4d shows the measured concentration profile of the NMC622 and NMC460 couple, while Figure S4 shows the SEM image and EDS mapping results near the interface. Figure 4d shows that the direction of the diffusion path of each element follows its initial concentration gradient. Table II summarizes the calculated diffusion constants for each diffusion couple discussed in Figure 4. The average effective diffusion coefficients, which can be understood as the flux per unit concentration gradient as described in the method section, of Mn, Ni and Co in the spherical CS particle (NMC622/NMC262) sintered at 850 • C (Figure 4b) are about 0.2 * 10 −16 , 0.49 * 10 −16 and −1.6 * 10 −16 m 2 /s, respectively and are about 1.6 * 10 −16 , 5.5 * 10 −16 and −37. * 10 −16 , respectively, for the pellet case heated at 900 • C (Figure 4c). The average effective diffusion coefficient of Ni is about three times of Mn in both cases. The effective diffusion co-efficient for Co is negative because Co moves opposite to the Ni motion. The absolute values of the diffusion coefficients in the spherical particles (4b) are much smaller than those in the pellet (4c), which can be attributed to higher temperature and different contacts between the core and shell materials in the pellet case (4c). The value of the average effective diffusion coefficient of Co is less meaningful in both cases (4b and 4c) since denominator is close to 0 (difference in the concentration of Co at the far end of core and shell structures) as shown in Equation 5 in the methods section. The average effective diffusion coefficients of Mn, Ni and Co in the NMC622/NMC460 pellet case (refer to Figure 4d    . Cell voltage as a function of specific capacity for CS20-1 (a1), CS20-2 (a2) and CS20-3 (a3) prepared at 850 • C respectively. Cell voltage as a function of specific capacity for CS20-1 (b1), CS20-2 (b2) and CS20-3 (b3) prepared at 900 • C, respectively. Specific capacity as a function of cycle number for CS20-1 (c1), CS20-2 (c2) and CS20-3 (c3) prepared at 850 • C respectively. Specific capacity as a function of cycle number for CS20-1 (d1), CS20-2 (d2) and CS20-3 (d3) prepared at 900 • C respectively.
average effective diffusion coefficient of Mn is slightly higher than Ni. This is because of the interdiffusion between Mn and both Ni and Co due to a relatively high Mn concentration gradient. These results show that ternary diffusion is a very complex problem and has a strong dependence on the relative concentration of each element. Figures 5a and 5b show the cell voltage as a function of specific capacity for CS10-1, CS10-2 and CS10-3 prepared at 850 • C and 900 • C respectively. The cells were tested between 2.5-4.4 V for 4 cycles, then 2.5-4.6 V for 10 cycles, followed by 2.5-4.8 V for 1 cycle, and 2.5-4.6 V for another 10 cycles with a current of 10 mA/g at 30 • C. Figures 5a1, 5a2 and 5a3 show a clear oxygen release plateau at ∼4.5 V for the CS20 series heated at 850 • C regardless of the lithium content, which is attributed to the Li-rich and Mn-rich shell. 21,22,36 Conversely, Figures 5b1, 5b2 and 5b3 show this plateau is barely noticeable in the CS20 series prepared at 900 • C, further confirming that there is virtually no Li-rich and Mn-rich shell remaining in the CS20-900 series as discussed in Figure 3. Figures 5c and 5d show the specific capacity as a function of cycle number for CS20-1, CS20-2 and CS20-3 prepared at 850 • C and 900 • C respectively. The reversible capacities of CS20-850-1, CS20-850-2 and CS20-850-3 at 4.4 V are 190(2), 182(1) and 184(2) mAh/g respectively, while they are 196(2), 193(1) and 192(1) mAh/g for CS20-900-1, CS20-900-2 and CS20-900-3 respectively. The Li-rich and Mn-rich shell materials have low specific capacity at voltages below 4.4 V compared to the Ni-rich materials, hence, it is reasonable that the CS20-850 series show relatively lower capacities compared to the CS20-900 series, due to the presence of the Li-rich and Mn-rich shell. A reversible capacity of ∼215(5) mAh/g for cells charged to 4.6 V was observed in both the 850 and 900 • C series. After the one charge to 4.8 V, a relatively rapid capacity fade rate was observed for all the cells compared to the 3.0-4.6 V cycling before the 4.8 V charge, indicating there is little benefit to activating these lithium rich CS materials to 4.8 V. 22 Figures S5-S7 and S8-S10 in the supporting information show the cell voltage as a function of specific capacity and cycling performance for the C, CS10 and CS33 series prepared at 850 • C and 900 • C respectively. Figures 6a and 6b show a summary of the reversible capacity as a function of target lithium content for cells made of the core (C), CS10, CS20 and CS33 series synthesized at 850 and 900 • C, respectively, when cycled between 2.5-4.4 V. Figure 6a shows that C-850-3 (x = 0.06), CS10-850-3, CS20-850-3, and CS30-850-3 have reversible capacities of ∼200(3), 198(1), 185(2) and 167(3) mAh/g, respectively, whereas Figure 6b shows that C-900-3, CS10-900-3, CS20-900-3 and CS30-850-3 have reversible capacities of ∼200(5), 196(1), 192(2) and 182(3) mAh/g respectively. In general, the reversible capacity decreases with increases of the initial shell thickness at both temperatures, due to the lower capacity of the Mn-rich shell. However, CS20 and CS33 sintered at 900 • C show higher capacity than the same materials sintered at 850 • C because of a thinner Mn-rich shell created at higher temperature due to interdiffusion. Figures 6c and 6d show a summary of the reversible capacity as a function of target lithium content for cells made of the core (C), CS10, CS20 and CS33 series synthesized at 850 and 900 • C, respectively, when cycled between 2.5-4.6 V. Figure 6c demonstrates that C-850-3, CS10-850-3, CS20-850-3 and CS33-850-3 have reversible capacities of 219(3), 220(3), 215(2) and 208(1) mAh/g respectively, where CS33-850-3 shows a relatively lower capacity, indicating that the shell thickness needs to be optimized. Figure 6d shows that the C-900-5, CS10-900-5, CS20-900-5 and CS33-900-5 have similar reversible capacities of ∼218(3) mAh/g. Figures 6e and 6f show the capacity fade from the second cycle of the first 3-4.6 V charge to the last cycle (22 nd cycle) as a function of the target lithium content for cells made of the C, CS10, CS20 and CS33 series sintered at 850 and 900 • C respectively. Figures 6e  and 6f show that the percentage of the capacity fade decreases with increasing lithium content. Additionally, Figure 6e shows that the CS20-850 series have the least capacity fade compared to the others with the same target lithium content. The capacity fade of C-850-3, a b c d e f Figure 6. Summary of the reversible capacity as a function of target lithium content for cells made of the core (C), CS10, CS20 and CS33 series synthesized at 850 • C (a) and 900 • C (b), respectively, when cycled between 2.5-4.4 V. Summary of the reversible capacity as a function of target lithium content for cells made of the core (C), CS10, CS20 and CS33 series synthesized at 850 • C (c) and 900 • C (d), respectively, when cycled between 2.5-4.6 V. Capacity fade from the second cycle of the first 3-4.6 V charge to the last cycle (22 nd cycle) as a function of the target lithium content for cells made of the C, CS10, CS20 and CS33 series sintered at 850 • C (e) and 900 • C (f) respectively.
CS10-850-3, CS20-850-3 and S33-850-3 is about 6.7(3), 6.09(2), 4.04(1) and 4.7(3) % respectively, where CS33-850-3 shows more capacity fade compared to CS20-850-3. This result suggests that an optimized final shell thickness is required for the best performance. Figure 6f shows there is more capacity fade for the C, CS10 and CS20 series sintered at 900 • C compared to the same series made at 850 • C. This is possibly due to a higher Ni content on the surface of CS10-900 and CS20-900 because of interdiffusion. Moreover, the capacity fade of C-900-3, CS10-900-3, CS20-900-3 and CS33-900-3 is about 7.5(1), 6.8(1), 5.5(1) and 4.2(1) respectively, which decreases with increasing shell thickness. CS33-900-3 shows similar capacity fade compared to CS20-850-3. These results suggest that changing the sintering temperature will significantly affect the final shell thickness and the CS surface composition due to interdiffusion, which must be controlled for the best cell performance. Samples CS20-850-3 and CS33-900-3 were selected from the 24 synthesized samples for testing in full cell Li-ion coin cells with graphite as the counter electrode using two different electrolytes. The control electrolyte was 1 M LiPF 6 in 3:7 v:v ethylene carbonate (EC): diethylcarbonate (DEC). PES211 electrolyte is the control electrolyte plus 2% prop-1-ene-1,3-sultone (PES) + 1% methylene methane disulfonate + 1% tri(trimethylsilyl) phosphite (TTSPi). 7 The cells were tested between 2.8 and 4.5 V using a rate of C/5 followed by one cycle of C/20 every 20 cycles. Figure 7 shows the capacity of the cells as a function of cycle number. Figure 7a shows that the cells made with CS33-900-3 and CS20-850-3 using control electrolyte have capacity retentions of ∼75% and 80%, respectively, after 100 cycles. The rate capability of the cells become worse with increasing cycling by comparing the capacity of the C/20 and C/5 cycles every 20 cycles, which suggests an impedance increase due to parasitic reactions with the electrolyte. Conversely, Figure 7b shows CS33-900-3 and CS20-850 have better capacity retention with PES211 electrolyte, which are ∼85% and 90% respectively after 100 cycles. Additionally, there is no significant increase in the capacity difference between the C/20 and C/5 cycles every 10 cycles, indicating that the cell impedance is under control. It is necessary to combine both surface modifications and electrolyte modification for cells operating at high voltages in order to achieve both high energy density and long life-time.

Conclusions
In summary, the effect of interdiffusion of TM atoms between the core and shell materials on the electrochemical performance was systematically studied by varying the shell thickness and sintering temperature. It was observed in SEM/EDS mapping that there is almost no Mn-rich shell maintained for CS samples with an initial shell content less than 10% when sintered at 850 • C and less than 20% when sintered at 900 • C. The concentration profile across the center slice of a spherical CS particle was measured using STEM and the results were modeled based on Fick's law. The ternary interdiffusion constants were then determined and it was found that Co moved from the shell into the core even when there was no initial Co concentration gradient. This phenomenon was confirmed with the laminar pellets experiments, which also showed that the ternary diffusion coefficients have a strong dependence on the initial concentration of each element.
Electrochemical testing showed that the CS20-850 series has an prolonged oxygen release plateau at ∼4.5 V whereas it is not observed in the CS20-900 series which further confirms the disappearance of the Mn-rich shell when heated at high temperatures due to significant interdiffusion. Analysis of the electrochemical testing results of cells made from the 24 synthesized samples strongly suggests that optimizing the sintering temperature and the initial shell thickness is essential for the best cell performance due to interdiffusion. Combining CS samples with electrolyte additives allowed a full cell charged to 4.5 V to deliver both high energy density and long-life time. Full cell coin cells made of CS20-850-3 showed ∼90% capacity retention after cycling to 4.5 V for 100 cycles. This study of interdiffusion in a ternary system of NMC materials will help design next generation positive electrode materials.