Influence of Interlayer Cation Ordering on Na Transport in P2-Type Na0.67–xLiy Ni0.33–zMn0.67+zO2 for Sodium-Ion Batteries

P2-type Na2/3Ni1/3Mn2/3O2 (PNNMO) has been extensively studied because of its desirable electrochemical properties as a positive electrode for sodium-ion batteries. PNNMO exhibits intralayer transition-metal ordering of Ni and Mn and intralayer Na+/vacancy ordering. The Na+/vacancy ordering is often considered a major impediment to fast Na+ transport and can be affected by transition-metal ordering. We show by neutron/X-ray diffraction and density functional theory (DFT) calculations that Li doping (Na2/3Li0.05Ni1/3Mn2/3O2, LFN5) promotes ABC-type interplanar Ni/Mn ordering without disrupting the Na+/vacancy ordering and creates low-energy Li–Mn-coordinated diffusion pathways. A structure model is developed to quantitatively identify both the intralayer cation mixing and interlayer cationic stacking fault densities. Quasielastic neutron scattering reveals that the Na+ diffusivity in LFN5 is enhanced by an order of magnitude over PNNMO, increasing its capacity at a high current. Na2/3Ni1/4Mn3/4O2 (NM13) lacks Na+/vacancy ordering but has diffusivity comparable to that of LFN5. However, NM13 has the smallest capacity at a high current. The high site energy of Mn–Mn-coordinated Na compared to that of Ni–Mn and higher density of Mn–Mn-coordinated Na+ sites in NM13 disrupts the connectivity of low-energy Ni–Mn-coordinated diffusion pathways. These results suggest that the interlayer ordering can be tuned through the control of composition, which has an equal or greater impact on Na+ diffusion than the Na+/vacancy ordering.


Experimental Methods:
Synthesis: P2 Na0.67Ni0.33Mn0.67O2was prepared by a solid-state reaction using the Ni0.33Mn0.67(OH)2precursor with Na2CO3.The Ni0.33Mn0.67(OH)2precursor was prepared by a co-precipitation method adopted from the work by Dahn et al. 1 A 1.0M total solution of NiSO4 and MnSO4 was prepared from NiSO4-6H2O (99%, Acros) and MnSO4-H2O (99%+, Thermo Scientific) salts were pumped at a constant rate of 0.1g/min into a 250mL vessel initially charged with 125mL of 0.37M NH3-H2O (Fisher) solution.The reaction was held at a constant 60°C and kept under an argon atmosphere to prevent oxidation of the precipitate.pH was kept constant at 9.8±0.1 regulating the flow of 1M NaOH (98%, Fisher) solution with a valve operated by a pH controller (Cole Parmer pH/ORP 400).All solutions were prepared with ultrapure water (18.2MOhm-cm).The volume in the reactor was held constant by pumping excess volume out of the reactor.The reactor was continuously stirred at a constant rate by a magnetic stir bar.The reaction was allowed to proceed for at least 3 hours before collecting the precipitated product in an argon-purged filter funnel.The collected precipitate was thoroughly rinsed with 3L ultrapure water that was deoxygenated by boiling and then by actively bubbling argon for at least 5 minutes.The rinsed precipitate was then dried under vacuum at 70°C overnight before being transferred into an argon-filled glovebox for storage to prevent oxidation.The powder was lightly ground to pass through a 150um sieve prior to further use.The stoichiometric ratio of nickel to manganese was verified to be the designed 1:2 by energy dispersive x-ray spectroscopy.Ni0.25Mn0.75(OH)2was prepared by an identical method with appropriate modification to the quantity of the Ni and Mn sulfates.
The TM(OH)2 powder and a proportional amount of Na2CO3 (99.95% anhydrous, Acros) and Li2CO3 (99%, Sigma Aldrich) were combined in stoichiometric amounts totaling 1 gram and were thoroughly mixed 3 times for 2 minutes at 2000RPM using a planetary mixer (THINKY) with zirconia balls with manual intermediate stirring.A 5% molar excess of sodium was used to compensate for evaporation at high temperatures.Samples prepared for the NPD and QENS measurements used 7 Li enriched Li2CO3 (99%, Sigma Aldrich) to minimize the high neutron absorption of 6 Li.The mixed powders were formed into a 12.7mm diameter pellet with a hydraulic press (Carver) at 5 metric tons.The pelletized material was calcined in alumina crucibles by heating at 5°C/min to 850 °C and held for 20 h in the air using a muffle furnace (Neytech Vulcan 3-550A).The furnace and samples cooled naturally to 500 °C before quenching in air.The quenched pellets were transferred to the glovebox to avoid exposure to the moisture in the air.The pellets were lightly ground until they could pass through a 400 mesh (38um) sieve prior to electrode preparation.

Electrochemical Characterization:
All electrochemical testing was conducted with 1M NaPF6 in propylene carbonate as the electrolyte with glassy fiber separators against sodium metal.Electrodes were prepared from 80 wt% active material, 10 wt% carbon (C45, Timcal), and 10 wt% PVDF (from 8wt% PVDF in nmethyl pyrrolidone) that were dried overnight at 70°C.Typical active material loading was 2-3 mg/cm 2 .Galvanostatic cycling (for voltage profiles and rate study) was conducted in CR2032-type coinc cells using an Arbin battery cycler.Operando sXRD was conducted in coin cells with a Kapton window to allow x-ray transmission using a Maccor battery cycler.

Physical Characterization: sXRD:
The powders were loaded into 1mm Kapton capillaries and sealed with modeling clay immediately prior to measurement.Diffraction measurements were conducted at sector 11-ID-C of the Advanced Photon Source at Argonne National Laboratory (samples PNNMO, LFN5, LFN10, LFN20, LSN10) with x-ray wavelength of 0.1173 Å, and at sector 28-ID-2 of the National Synchrotron Light Source II (NSLS-II) at Brookhaven National Laboratory (sample NM13) with a x-ray wavelength of 0.1814 Å.The data collected at 11-ID-C are reproduced from our previous publication. 2Operando sXRD was conducted at beamline 17-BM of the Advanced Photon Source with a wavelength of 0.24145 Å.

NPD:
The powders (~0.2g) were loaded into 3mm quartz capillaries and sealed with epoxy inside an argon-filled glovebox (<0.5ppmO2) prior to measurement at the Nanoscale-Ordered Materials Diffractometer (NOMAD, BL-1B) instrument at the Spallation Neutron Source facility at Oak Ridge National Laboratory. 3 Two 24 min scans were collected on each powder sample and then summed to improve the counting statistic.Scattering signal from the empty vanadium can measurements was subtracted as background from the sample measurement and data were normalized by the scattering intensity from a 6 mm vanadium rod to correct for detector efficiencies.
ICP-MS: ICP-MS was conducted at the Boise State University Isotope Geology Laboratory to determine the overall composition.The samples were dissolved in 2.5 mL 16M HNO3 + 2.5mL H2O and 100 µL 30% H2O2, diluted gravimetrically in 2% HNO3 for analysis.The solution was analyzed by a ThermoScientific, iCAP-RQ, inductively coupled plasma mass spectrometer.

QENS:
Quasielastic neutron scattering measurements were conducted at the Backscattering Silicon Spectrometer (BASIS, BL-2) instrument at the Spallation Neutron Source of Oak Ridge National Laboratory. 4 Samples were heated in an argon glovebox to at least 200°C before transfer into the sample canister to remove any adsorbed water.The samples were sealed in an annular aluminum can with a 2 mm gap for the measurement.In this experiment, BASIS was operated at 60 Hz chopper frequency with incident neutron of bandwidth centered at 6. 4 Å.This instrument configuration provides a Q range of 0.2 Å -1 to 2.0 Å -1 with an energy transfer range of 100 µeV and an energy resolution of 3.6 µev.Short elastic scattering intensity measurements as a function of temperature was performed from 30 K to 680 K at a 1K/min heating rate.Long QENS spectra were collected from each sample at 450 K, 525 K, 600 K, and 675 K.The temperatures were maintained using closed cycle refrigerator with helium exchange gas.Sample specific instrument resolution function was measured at 30 K. Data from an empty can at 675 K were collected and subtracted as a background.Standard vanadium dada was used for detector efficiency normalization.QENS data were reduced in the Mantid package 5 and analyzed using QClimax software. 6The QENS intensity signal, I (Q,E), were fitted with a one component model as: where  1 (𝑄) captures the elastic scattering fraction.The delta function, ( ) E  , which is centered at zero energy transfer, accounts the elastic contribution.The dynamic structure factor, ( )

Simulation and Refinement of the NPD Patterns:
The refinement of the (clustered) faulting model was conducted in FAULTS (based on DiFFAX) 7,8 , and traditional Rietveld refinement was performed with GSAS-II. 9In both cases, the isotropic displacement factors (Biso = 8π 2 Uiso), the oxygen positions, and Na site occupancies were constrained to be equivalent between each phase/clustered domain with appropriate modifications made for the different unit cells.The total Na site occupancy was constrained to require the sum of all site occupancies correspond to 0.67Na per formula unit, and the Ni/Mn site occupancy was constrained to match the designed ratio by adjusting the occupancy of Ni on the A site (NiA) with an appropriate amount of Mn.Due to the similar scattering behavior of Li and Mn for neutrons (bcoh,Mn = -3.73fm, bcoh,Li = -1.9),no adjustment was made for LFN5 to reduce correlation between the refined parameters (MnNi especially).The oxygen x and y positions were constrained to allow an expansion of the Ni octahedron and equivalent contraction of the Mn octahedra that preserves a single but distinct bond length for each TM species (Ni-O ~ 2.04 Å, Mn-O ~ 1.91 Å).For the ABAB ordered structure, this reduces the TM-ordered structure (space group: P63) to monoclinic symmetry (space group: C2, a'=√3a, b=3a, c'=c, β=90.0°), while the ABCABC ordered structure maintains its symmetry (space group: R3 ̅ c) but with oxygen x-coordinate allowed to deviate from 1/3.Though not used for any refinements, the AAAA structure can maintain the P63 space group with similar deviation of the oxygen x-coordinates if Ni is located at the origin.
The GSAS-II refinement used the instrument profile parameters provided by the NOMAD (SNS BL1-B) beamline staff from a fit to SRM 640e Si powder.Due to the software limitation of FAULTS to only handle constant wavelength diffraction data, time-of-flight diffraction data cannot be processed directly in FAULTS.The time-of-flight data was first converted to dspacing according to the instrument parameters difA, difC, and Zero, and time-of-flight (TOF) according to the equation d = [-difC + (difC 2 -4difA(Zero-TOF))]/(2difA).The data were then converted from d-spacing to a pseudo-constant wavelength (pCW) form according to Bragg's law by assuming an artificial wavelength (1.54A) near the center of the instrument's wavelength range.The "pseudo-θ" (θ') derived from this process was used as the x-coordinate for the refinement with FAULTS and included the data from d = 1.0 -6.8Å.The sensitivity of the refined parameters to the peak shape requires consideration of the difference in profile functions between time-of-flight neutron diffraction and constant-wavelength diffraction.The TOF peak profile consists of a Gaussian component convoluted with two back-to-back exponentials, where the full-width at half maximum (FWHM) Gaussian component (σ) is a function of d-spacing according to σ(d) = σ0 + σ1d 2 + σ0d 4 .The Gaussian part of the pCW peak profile is approximated by a pseudo-Voigt function 10 with FWHM that depends on θ according to Γ(θ) =

Figure S1 .
Figure S1.Simulated NPD patterns with different modes of Ni/Mn site mixing.If Ni and Mn are mixed only between two sites (Ni on A, Mn on B or C), (hkl) dependent superlattice peak intensity changes occur, while equal mixing on all sites reduces the intensity of all superlattice peaks similarly.In any case, mixing cannot fully suppress the intensity of the (102) peak at 3.42Å.

Figure S2 :
Figure S2: The effect of the domain size (1/Pt) for an equal mixture (50% layer fractions) of ideally ordered ABAB and ABCABC domains.

Figure S4 .
Figure S4.The XRD pattern of NM13 without background subtraction.

Figure S5 .
Figure S5.Operando sXRD of LFN10 during charge/discharge between 2-4.2V.Peaks are indexed to the typical P63/mmc unit cell (a x a x c).

Figure S11 .
Figure S11.Schematic of the disruption caused by the higher Mn content in NM13 compared to PNNMO, where the red crosses indicate Ni-Mn sites that would be replaced by Mn-Mn sites.The blue isosurfaces are those calculated for PNNMO, which would be blocked at the sites indicated by the crosses.

Figure S12 .
Figure S12.Comparison of the calculated diffraction patterns with equivalent parameters other than whether faults may occur on all layers (as in simulations) or every other layer (necessary for refinement).

Figure S13 .
Figure S13.Optimized crystal structures for calculating site energies with a Na16Li1Mn16Ni7O48 structure.The Na1, Na2, and Na3 atoms are at Mn-Li, Mn-Mn, and Mn-Ni coordination.

2 , 2 1+𝐷𝑄 2
contains the information about the mobility of Na ions in the samples.A data at each Q were fitted after adding a linear background, (B(Q,E)) and a convolution (  ) with the instrument resolution function, ( )E Q R , .Asingle Lorentzian function was used to model ( )  is the half width at half maximum (HWHM) of quasi elastic signal.Q-dependence of  was further analyzed using a jump diffusion model, () =    , from which the diffusion coefficient (D) and the residence time ( o  ) are obtained.These parameters can be used to obtain a jump length (L) as,  2 = 6 0 .

Table S1 .
ICP-MS results for the materials (normalizing the sum of the transition metals to 1.00):