The Limited Incorporation and Role of Fluorine in Mn-rich Disordered Rocksalt Cathodes

Disordered rocksalt oxide (DRX) cathodes are promising candidates for next-generation Co- and Ni-free Li-ion batteries. While fluorine substitution for oxygen has been explored as an avenue to enhance their performance, the amount of fluorine incorporated into the DRX structure is particularly challenging to quantify and impedes our ability to relate fluorination to electrochemical performance. Herein, an experimental–computational method combining 7Li and 19F solid-state nuclear magnetic resonance, and ab initio cluster expansion Monte Carlo simulations, is developed to determine the composition of DRX oxyfluorides. Using this method, the synthesis of Mn- and Ti-containing DRX via standard high temperature sintering and microwave heating is optimized. Further, the upper fluorination limit attainable using each of these two synthesis routes is established for various Mn-rich DRX compounds. A comparison of their electrochemical performance reveals that the capacity and capacity retention mostly depend on the Mn content, while fluorination plays a secondary role.


Density functional theory calculations
All the DFT calculations were performed using the Vienna ab initio simulation package (VASP) using the projector-augmented wave method, 1 a plane-wave basis set with an energy cutoff equal to 680 eV, and a reciprocal space discretization of 25 k-points per Å−1 .All the calculations were converged to 10 −6 eV in total energy for electronic loops and 0.02 eV/ Å in interatomic forces for ionic loops.We relied on the regularized strongly constrained and appropriately normed meta-GGA exchange-correlation functional (r 2 SCAN), 2 which is believed to better capture cation-anion hybridization and Li-coordination.r 2 SCAN has better computational efficiency performance than the earlier version of SCAN. 3

Monte Carlo cluster expansions
The cluster-expansion (CE) technique was used to study the configurational thermodynamics of materials in which sites can be occupied by multiple cations and has been applied 6][7] The CE expands the energy of multicomponent disordered rocksalt materials as a sum of many-body configurational interactions: where σ i is the occupancy of different species (indicator basis site function) and J refers to the effective cluster interaction (ECI) using the sinusoid basis. 4For the simulation of atomic orderings, a cluster-expansion Hamiltonian was generated in the chemical space of Li + -Mn 3+ -Mn 4+ -Ti 4+ -O 2− -F − , with pair interactions up to 7.1 Å, triplet interactions up to 4.0 Å, and quadruplet interactions up to 4.0 Å based on a primitive cell of the rocksalt structure with lattice parameter a = 3 Å.In total, 162 ECIs (including the constant term J 0 ) were defined, and the CE Hamiltonian was fitted with 563 different structures.As the CE Hamiltonian was defined on a high-dimensional multicomponent system, the ECIs were fitted using the appropriate method to address the complexity-induced over-fitting. 8,9The ECIs were determined with the optimal sparseness and cross-validation error (< 8 meV/atom) with a ℓ 0 ℓ 2 -norm regularized regression. 10We refer readers to Ref. 11 for details of CE in ionic systems.To simulate atomic orderings at equilibrium, we used canonical Monte Carlo simulation with the Metropolis-Hastings algorithm.Overall, 1,000 representative structures (640 atoms per structure) were sampled from the equilibrium ensemble.We used smol for the CEMC simulations 12 and pymatgen for the structure processing. 13

Materials synthesis
Solid-state synthesis of DRX A standard solid-state synthesis protocol was used to synthesize LMT53, LMT62, and LMT81.Stoichiometric amounts of Li 2 CO 3 , LiF, Mn 2 O 3 , and TiO 2 precursors were used, with 10% Li excess to compensate for possible Li loss.Precursor powders were mixed via wet ball-milling with ethanol in a planetary ball-mill at 300 rpm for 6 hours.The resulting slurry was dried to form a powder, and pressed into 200 mg pellets.The pellets were then heated at targeted temperatures for specified reaction times under Ar gas flow in a tube furnace, after which the furnace hood was opened and pellets were allowed to cool naturally.

Microwave synthesis of DRX
For microwave synthesis of LMT53, LMT62, and LMT81, stoichiometric amounts of Li 2 CO 3 , LiF, Mn 2 O 3 , and TiO 2 precursors were used; no Li excess was added.Precursors were mixed and pelletized according to the same procedure used for solid-state synthesis, where precursor powders were ball-milled with ethanol at 300 rpm for 6 hours, dried, and then pressed into 200 mg pellets.A double crucible setup was used for microwave synthesis, where a small alumina crucible was placed inside a larger crucible filled with 5 g of activated charcoal.
The precursor pellet was placed in the small alumina crucible on top of a layer of sacrificial precursor powder to prevent a reaction between the alumina surface and the pellet.The entire setup was placed in a conventional 1200W microwave, and heated at appropriate times and microwave power in an ambient air atmosphere.Upon termination of microwaves, the pellet was immediately quenched into a beaker of distilled water to stabilize the DRX phase.The DRX pellet was then dried on a hot plate and ground into powder.Optimized microwave power values for LMT44, LMT53, and LMT62 were 600W, 480W, and 480W, respectively, and optimal synthesis times were 5 min for all compositions.

Characterization
Powder X-ray diffraction data for all samples were collected using a laboratory-source Pananlytical Empyrean diffractometer with Cu Kα radiation in reflection geometry.The TOPAS software suite was used for Rietveld refinements of data sets.were processed using Bruker TopSpin 3.6.0and spectra were fit using an in-house developed python script.In the first step, the as-synthesized sample is examined using powder XRD to identify potential crystalline impurity phases.Samples that contain transition metal oxide or oxyfluoride phases other than the disordered rocksalt phase of interest are discarded, as are samples containing a significant amount of crystalline impurities, such as LiF, Li 2 CO 3 , or Li 2 O.

Electrochemistry
For samples that are (close to) phase pure by XRD, the composition of the sample is analyzed in a second step using ICP and a F-ISE, which provide the overall ratio of cationic species and the F content, respectively.Combining XRD, ICP, and F-ISE results, it is clear that all transition metal species are integrated into the DRX structure, but the presence of (at least partially) amorphous Li-and F-containing impurity phases means that such phases must be quantified in order to obtain the Li and F contents in the DRX phase.
Li-containing impurity phases (crystalline and/or amorphous) can be quantified using 7 Li ssNMR.A typical 7 Li spectrum obtained on a DRX sample is shown in Figure S2, or from diamagnetic domains in the DRX phase. 15However, for the compositions of interest to this work, the probability of forming diamagnetic Li environments in the DRX structure is negligible due to the high concentration of paramagnetic metal species (here, redox-active Mn), and the diamagnetic signal can therefore be assigned to Li-containing impurity phases.The experimentally determined fraction of F in the DRX, F exp , is calculated by dividing the integrated paramagnetic signal intensity p by the total signal intensity, which is the sum of p and the integrated diamagnetic signal intensity d (see Figure S2).
Given that p is underestimated experimentally, F exp provides a lower bound for the amount of F in the DRX structure.An adjusted (quantitative) paramagnetic intensity, P , and a scaled DRX F fraction, F scaled , can be determined if the fraction of NMR-visible F environments in the DRX phase is known.The fraction of NMR-visible F sites is hereafter referred to as F-Mn(0), as none of the six nearest-neighbor cations (C) surrounding those F species can be Mn.The adjusted paramagnetic intensity P is calculated using the equation P = p/F-Mn(0).
Subsequently, F scaled can be obtained using:  Figure S3: Fits of 7 Li ssNMR spectra for optimized solid-state and microwave synthesized LMT53, LMT62, and LMT81 samples.ssNMR spectra were acquired at 2.35 T with a magic angle spinning (MAS) speed of 60 kHz with a long 20 sec recycle delay to ensure sufficient relaxation of diamagnetic environments.Fits were performed using an in-house developed python program.For each DRX sample, an initial fit was first done on a ssNMR spectrum acquired with a short recycle delay of 50ms to obtain high quality fits to the paramagnetic lineshape (see Figure S2).The model obtained from this fit then used for the final fitting on sufficiently relaxed spectra with a 20 sec recycle delay (shown in this figure ), where only component peak widths and magnitudes were allowed to vary.Diamagnetic impurity percentages were extracted by integrating paramagnetic and diamagnetic components of the fit ssNMR spectra, and results for each sample are listed in Table S2.S3.
Table S1: Chemical formulae of the DRX supercells and temperature inputs for the ab initio CEMC simulations used to determine the distribution of F environments and F content in the LMT53, LMT62, and LMT81 samples of interest.For the solid-state synthesized samples, the Mn oxidation state was assumed to be 3+, while for the microwave synthesized samples, a mixture of 3+ and 4+ Mn oxidation states was assumed.For solid-state synthesized samples, CEMC simulation temperatures were chosen to be as close as possible to the experimental sintering temperatures.For microwave synthesized samples, a 1600 • C CEMC simulation temperature was chosen throughout since the temperature of the DRX pellets was found to be 1200 • C a few seconds after microwave heating, suggesting a higher heating temperature closer to 1500-1600 Figure S5: a) Frequency of F species with no nearest-neighbor Mn, F-Mn(0), obtained from ab initio CEMC simulations on various DRX compositions (including formulae with 5% Mn 4+ ) and at various temperatures.b) F-Mn(0) plotted against the fraction of cation sites occupied by Mn in the DRX structure at various simulation temperatures.
Table S1.For samples sintered at 1100 • C, the Li and F contents shown in Figure S7 correspond to the bulk sample composition (as obtained from ICP and F-ISE measurements) and not to the DRX phase itself.Since those Li and F contents are already drastically reduced compared to samples sintered at lower temperatures, clearly indicating that those conditions are not optimal for DRX fluorination, we did not deem further investigation using ssNMR and CEMC simulations necessary.For all DRX compositions, similar trends were observed regarding the impact of reaction time and temperature on F incorporation into the DRX phase and on the impurity content.
Regardless of whether a 2h or 12h calcination step was used, high reaction temperatures (1100 • C) resulted in a drastically reduced amount of F and Li in the DRX phase and sample (Figures S7a-b), which is consistent with increased LiF vaporization.As seen in Figures S7cd, a longer reaction time of 12h is desired when optimizing for phase purity, as the amount of Li and F impurities in the DRX samples is reduced, particularly for the LMT53 and LMT62 compositions.Interestingly, greater fluorination of the DRX phase can be observed after a 12h calcination step for LMT53 and LMT62, as shown in Figure S7a, indicating that longer reaction times at temperatures of 800 or 900 • C favor F incorporation.In contrast, a very low amount of F is able to incorporate the DRX structure for LMT81 (only 1-1.5% fluorination), irrespective of the sintering time.
Overall, the results presented in this section indicate that lower reaction temperatures and longer reaction times result in greater fluorination of the DRX phase, as well as greater volatilization of (at least partially amorphous) F-and Li-based impurities.The DRX stoichiometries of LMT53, LMT62, and LMT81 samples prepared in this way (using the minimum temperature needed to obtain a phase-pure DRX and a 12h calcination step) are listed in Table S2, while the amount of F that is present as LiF or has volatilized during the synthesis is provided (in mols/ pfu of DRX) in Table ??.
Microwave synthesis.While the precursors used for microwave synthesis were similar to those used for solid-state synthesis, no Li excess was used with this synthesis route.In fact, our recent work on microwave synthesized Li 1.2 Mn 0.4 Ti 0.4 O 2 showed that the addition of excess Li results in significant Li-containing impurities in the sample due to a very small amount of Li volatility during the 5 minute heating process. 18e only two microwave synthesis parameters that can be controlled with our current setup are reaction time and microwave power.As the reaction temperature depends on both of these factors, the reaction time was held constant at 5 minutes to ensure the formation of a phase-pure product while avoiding melting of the precursor pellet, and only the microwave power was varied.Two extremes of microwave power were chosen: the maximum available power of 1200W, and the minimum power needed to consistently produce a phase-pure DRX within 5 minutes, corresponding to 720W, 600W, and 600W for mw-LMT53, mw-LMT62, and mw-LMT81, respectively (Figure S8).Although the lower power needed to synthesize higher Mn content DRX may seem counter-intuitive (considering that the solid-state synthesis results indicated that higher temperatures were needed to obtain those systems), we note that Mn-rich oxides appear to couple more strongly to the microwaves, resulting in more effective heating as evidenced by the fact that the mw-LMT81 pellet glows even at low microwave powers.
Figure S8: XRD patterns collected on LMT53, LMT62, and LMT81 samples prepared via microwave synthesis using various microwave powers and a 5 minute reaction time.

Supplementary Note 1 :
Electrochemical performance of DRX was tested in coin cells assembled in an Ar filled glovebox.As-synthesized DRX powders were first carbon coated and downsized by ball milling with Super C65 at 400 rpm for 6h in a planetary ball mill.DRX cathode films were then produced by mixing the post-processed, carbon coated DRX powders with polytetrafluoroethylene (PTFE) such that the active material: carbon: binder ratio was 70:20:10.Films were then rolled out and punched to form 6.35 mm diameter discs with loading densities of 5-6 mg/cm 2 .CR2023-type coin cells were assembled using the cathode film and a Li metal anode.1M LiPF 6 in ethylene carbonate and dimethyl carbonate (EC/DMC with a 1:1 volume ratio) electrolyte was used, along with a glass fiber (Whatman GF/D) separator.Coin cells were cycled in a temperature controlled chamber (Neware MHW-25-S) set to 25 • C using a Biologic VMP-3 tester.Hybrid experimental-computational methodology to determine DRX stoichiometry Does fluorination impact DRX?The role of fluorine in DRX cathode A flowchart summarizing the methodology used to determine DRX stoichiometries is shown in Figure S1.The method described below builds upon and extends the experimental approach recently proposed by some of us 15 to determine the composition of DRX samples, and improves the accuracy of F content predictions.

Figure S1 :
Figure S1: Flowchart of process to determine DRX stoichiometry.
where two types of resonances are observed: a sharp resonance centered around 0 ppm corresponding to Li in diamagnetic environments (red dotted line), and broader resonances associated with paramagnetic Li environments (blue dotted lines).The diamagnetic signal either results from diamagnetic impurity phases in the sample, such as Li 2 CO 3 and Li 2 O,16 The broad paramagnetic signals can be attributed to Li species in the DRX phase, which are in close proximity to the redox-active metals (referred to as M hereafter), resulting in strong paramagnetic couplings between the7 Li nuclear spin and unpaired electron spins originating from nearby M d orbitals.Those paramagnetic interactions manifest as a large Fermi contact shift and significant broadening of the resonance.Additionally, the extensive disorder of cationic and anionic species in DRX compounds leads to a large number of possible Li local environments, each with a (slightly) different shift, and extensive overlap of the resulting broad paramagnetic resonances.By fitting the sharp diamagnetic signal and broad paramagnetic resonances (Figure S2), the molar fraction of Li in impurity phases can be determined and used to scale ICP results and derive the Li stoichiometry of the DRX phase.Similarly to7 Li ssNMR, typical 19 F ssNMR spectra obtained on DRX cathode samples consist of broad paramagnetic and sharp diamagnetic resonances.In the case of 19 F ssNMR, the sharp diamagnetic peak at -204 ppm is attributed to LiF,17 and the broader features correspond to F species in paramagnetic (DRX) environments.Unlike Li, F can be directly bonded to paramagnetic M species which, compounded with the high gyromagnetic ratio (γ) of 19 F, results in extremely strong paramagnetic interactions and extremely fast 19 F ssNMR signal decay.Thus, F species bonded to at least one paramagnetic Mn ion are effectively invisible by NMR, and integration of the 19 F ssNMR signals underestimates the amount of F in DRX environments.In order to quantify the amount of F in the DRX structure from 19 F ssNMR, the fraction of "NMR (in)visible" F environments must be determined.Here, ab initio cluster expansion Monte Carlo (CEMC) simulations allow us to predict the equilibrium distribution of F environments in the DRX compounds of interest at typical synthesis temperatures.From this, the integrated paramagnetic 19 F ssNMR signal intensity is scaled to account for signal loss, and the DRX F content is obtained by combining the F-ISE and 19 F ssNMR results, as discussed in Supplementary Note 2.Supplementary Note 2: Analysis of the 19 F ssNMR results to quantify the F content in the DRX phase As mentioned in Supplementary Note 1, while the integration of 19 F ssNMR signals should, in theory, provide quantitative insight into the distribution of F species in the DRX phase and in impurity phases, F species directly bonded to Mn cannot readily be observed, resulting in an underestimation of DRX fluorination.Two possible approaches to scale the integrated intensity of the paramagnetic 19 F ssNMR signal (p) are presented below, which either rely on simple probability calculations, or on CEMC simulations of the distribution of F environments in the structure.
Here, we use CEMC simulations of a 640-atom supercell to model the distribution of environments (and any short-range ordering) in LMT53, LMT62, and LMT81.Direct counting of the F environments provides F-Mn(0) at the simulation temperature, from which the fraction of F in the DRX can be obtained by scaling the integrated paramagnetic signal intensity.The DRX fluorination level is then calculated by scaling the total F content in the DRX sample (obtained from F-ISE measurements) by F scaled .If CEMC simulations are not available, an upper bound for the amount of F in the DRX structure can be approximated by computing F-Mn(0) assuming a random distribution of species within the DRX structure.The assumption of complete disorder leads to an overestimation of the number of Mn-F bonds, and thereby of the fraction of NMR invisible F environments, and to an upper bound for the DRX F content.15

Figure S2: Representative fit of the 7
Figure S2: Representative fit of the 7 Li ssNMR spectrum collected on ss-LMT62 with a short 50ms recycle delay, illustrating the broad paramagnetic and sharp diamagnetic signals corresponding to Li in the DRX and in Li-containing impurity phases, respectively.

Figure S4 :
FigureS4: Fits of 19 F ssNMR spectra for optimized solid-state and microwave synthesized LMT53, LMT62, and LMT81 samples.ssNMR spectra were acquired at 2.35 T with a magic angle spinning (MAS) speed of 60 kHz.All spectra shown except for ss-LMT81 were acquired with a long 20 sec recycle delay to ensure sufficient relaxation of diamagnetic environments.For ss-LMT81, the spectrum shown was acquired with a 50 ms recycle delay, as no 19 F ssNMR signal was observed when a 20 cycle recycle delay was used, due to the almost negligible fluorine content in the sample.Semiquantitative fluorine impurity percentages obtained from these fits are listed in TableS3.

Figure S7 :
Figure S7: Compositional analysis of LMT53, LMT62, and LMT81 samples prepared via solid-state synthesis using various sintering temperatures and times.(a) Li and (b) F content in the DRX phase for samples synthesized at 800, 900, and 1000 • C, and in the bulk sample for those prepared at 1100 • C. Amount of (c) Li-and (d) F-containing impurities in samples synthesized at 800, 900, and 1000 • C. Error bars in (c) correspond to the upper and lower bounds for the amount of F-containing impurities obtained from 19 F ssNMR and assuming a random distribution of species in the DRX structure (see Supplementary Note 2 for details of the analysis).

Figure S9 :
Figure S9: Compositional analysis of LMT53, LMT62, and LMT81 samples prepared via microwave synthesis using various microwave powers and a 5 minute reaction time.(a) Li and (b) F content in the DRX phase, and (c-d) amount of (c) Li-and (d) F-containing impurities in the as-prepared samples.Error bars in (c) correspond to the upper and lower bounds for the amount of F-containing impurities obtained from 19 F ssNMR and assuming a random distribution of species in the DRX structure (see Supplementary Note 2 for details of the analysis).

Figure S10 :
Figure S10: SEM images of solid-state and microwave synthesized LMT53, LMT62, and LMT81, after a ball milling step with Super C65 to downsize and carbon coat active materials.

Figure S13 :
Figure S13: Plots showing the (a) average discharge voltage, (b) energy density, (c) voltage hysteresis (the difference between average charge and discharge voltages), and (d) Coulombic efficiency as a function of cycle index, for solid-state and microwave synthesized LMT53, LMT62, and LMT81.

Table S2 :
DRX stochiometries of all optimized microwave and solid-state synthesized LMT53, LMT62, and LMT81 samples.While the listed F content is the true F stoichiometry in the DRX phase obtained by combining F-ISE, 19 F ssNMR and ab initio CEMC results, F min and F max correspond to the minimum and maximum possible DRX F contents obtained from 19 F ssNMR and assuming a random distribution of species in the DRX structure (see Supplementary Note 2 for details of the analysis).Li bulk corresponds to the total Li content in the DRX sample including impurities, as measured by ICP.The amount of Li impurity (in mol %) refers the molar fraction of Li present as diamagnetic species in the sample.