Improved cycling stability in high-capacity Li-rich vanadium containing disordered rock salt oxyfluoride cathodes

af Lithium-rich transition metal disordered rock salt (DRS) oxy ﬂ uorides have the potential to lessen one large bottleneck for lithium ion batteries by improving the cathode capacity. However, irreversible reactions at the electrode/electrolyte interface have so far led to fast capacity fading during electrochemical cycling. Here, we report the synthesis of two new Li-rich transition metal oxy ﬂ uorides Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F using the mechanochemical ball milling procedure. Both materials show substantially improved cycling stability compared to Li 2 VO 2 F. Rietveld re ﬁ nements of synchrotron X-ray di ﬀ raction patterns reveal the DRS structure of the materials. Based on density functional theory (DFT) calculations, we demonstrate that substitution of V 3+ with Ti 3+ and Fe 3+ favors disordering of the mixed metastable DRS oxy ﬂ uoride phase. Hard X-ray photoelectron spectroscopy shows that the substitution stabilizes the active material electrode particle surface and increases the reversibility of the V 3+ /V 5+ redox couple. This work presents a strategy for stabilization of the DRS structure leading to improved electrochemical cyclability of the materials.


Introduction
Lithium-ion batteries (LIBs) are the most widely used energy storage systems for applications in electric transportation and portable electronic devices. 1,2 Cathode materials for LIBs need to meet high energy and power density criteria for several applications and are the main focus of current research activities. 1,3,4 In 2015, the lithium-rich disordered rock salt (DRS) Li 2 VO 2 F was identied as a promising cathode material with a high theoretical capacity of 462 mA h g À1 by Chen et al. 5 This compound exhibits a DRS structure where Li and V ions appear to be randomly distributed at the same crystallographic site in the crystal structure. 6 The lithium-excess (Li : TM-ratio (transition metal) > 1) in DRS structures enables percolating pathways for Li-ions to diffuse through the structure, facilitating their use as battery materials with relatively high capacities. 7,8 Li 2 VO 2 F was the rst material reported in the class of Li-rich DRS materials and the rst material to incorporate F À by partly substituting O 2À to lower the oxidation state of the transition metal cation and to increase the average discharge potential. 9 The capacity is expected to originate from the two-electron redox couple, V 3+ /V 5+ , leading to practical capacities over 300 mA h g À1 , considerably higher than what is observed for conventional cathode materials such as LiCoO 2 and LiFePO 4 (140 mA h g À1 and 170 mA h g À1 ). [1][2][3] Many different compounds with the Li-rich DRS phase have been reported since the original study. Materials like Li 1.3 Nb 0.3 TM 0.4 O 2 (TM ¼ Fe 3+ , Mn 3+ , V 3+ ) and Li 1.2 Ni 1/3 Ti 1/3 Mo 2/15 O 2 have been synthesized using a classical high-temperature annealing solid-state approach. [10][11][12] These materials exhibit a relatively stable electrochemical behavior for more than 20 cycles. Fluoride-containing Li-rich DRS materials, such as Li 2 VO 2 F, Li 2 Cr 1Àx V x O 2 F (x ¼ 0-1), Li 2 MnO 2 F, Li 2 MoO 2 F and Li 2.1 Ti 0.2 Mo 0.7 O 2 F, have on the other hand only been synthesized by high-energy ball milling, leading to nano-sized particles with crystallographic defects. 5,[13][14][15][16][17] Ball milled DRS oxyuorides generally suffer from fast capacity fading and the reason for the capacity fading is still not fully understood. 5,15,16,18 Recent publications indicate that such DRS phases might be metastable. 10,14,19 A recent study of Källquist et al. identied irreversible reactions at the surface of Li 2 VO 2 F with the electrolyte as the possible cause of the capacity fading. 20 Herein, we propose that the key to understand the degradation mechanism of Li 2 VO 2 F and other disordered rock salt oxyuorides is linked to the metastable structure and surface stability of the DRS phase. Even though Li 2 VO 2 F has a lower average discharge voltage compared to other DRS materials, its high theoretical and practical discharge capacities make it an attractive material to study. We report the synthesis of two new Li-rich DRS oxy-uoride phases, Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F. We also elucidate alternative synthesis approaches towards a conventional solid-state synthesis of Li 2 VO 2 F and substituted Li-rich DRS oxyuorides (see ESI †). LiF cannot be incorporated stoichiometrically into the structure under conventional solidstate synthesis conditions. Thus, the isovalent substituted Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F compounds are synthesized by a high energy ball milling approach and the properties of the compounds are compared with those of Li 2 VO 2 F. The structural stability of the materials is investigated by density functional theory (DFT) calculations. Further, we analyze the electrochemical performance by galvanostatic charge-discharge experiments and atomic level information from the material surface is obtained from ex situ hard X-ray photoelectron spectroscopy (HAXPES).

Synthesis procedures
The compounds were prepared by using a dry ball milling procedure in two steps (each step involves milling at 600 rpm for 20 h using a Fritsch Pulverisette 6 classic line containing an 80 mL Si 3 N 4 jar and 25 balls of 10 mm diameter under the conditions of a ball to powder ratio of 10 : 1 and the total amount of powder of approx. 4.5 g). All compounds were added into an air-tight Si 3 N 4 jar under inert conditions in an argon-lled glovebox with water and oxygen levels below 0.1 ppm.
In the rst step, Li 2 O (10% excess to compensate loss during the synthesis, 21  Powder X-ray diffraction (PXRD) Synchrotron PXRD patterns were recorded at beamline 11-ID-B at the APS Argonne National Lab using a PerkinElmer at panel detector (XRD1621) with a pixel size of 200 Â 200 mm. The powdered samples were placed in 1.1 mm borosilicate glass capillaries inside an argon-lled glovebox and sealed before they were measured in transmission geometry. A sample-todetector distance of 641.010 mm and a wavelength of 0.21280 A were used for data acquisition with a summed exposure time of 30 s per diffraction pattern. The diffraction data were integrated using the Fit2D. [22][23][24] Rietveld renements were conducted using TOPAS version 5. 25 The instrumental resolution function was determined using the CeO 2 standard. For the Rietveld renement, the b-factors were rened for the 4a and 4b Wyckoff positions and not for each element individually. The occupancies of O and F were xed to 2/3 for O and 1/3 for F, respectively. The TM and Li occupancies were restrained to result in 1 and rened.
Ex situ PXRD patterns of cycled electrodes aer 100 cycles were recorded in reection geometry using a STOE STADI-p diffractometer with Mo K a1 radiation (0.70932Å), equipped with a DECTRIS MYTHEN 1K strip detector. The electrodes where covered with Kapton® tape to prevent oxidation in air.

Transmission electron microscopy (TEM)
TEM characterization and energy dispersive X-ray spectroscopy (EDX) mapping were performed using a Cs corrected JEOL ARM CF operating at 200 kV, equipped with an SSD Jeol EDX spectrometer.

DFT calculations
DFT calculations were carried out using the Vienna Ab initio Simulation Package (VASP) [26][27][28][29] using the projector augmentedwave (PAW) method. 30 The generalized gradient approximation as parametrized by Perdew, Burke and Ernzerhof 31 was used as the exchange-correlation functional. The plane-wave cutoff of 600 eV was used, and both the cell and atomic positions were fully relaxed such that all the forces are smaller than 0.02 eV A À1 . A rotationally invariant Hubbard U correction 32,33 was applied to the d orbitals of V, Ti and Fe with the U values of 3.25, 3.50 and 4.30 eV, respectively. Integrations over the Brillouin zone were carried out using the Monkhorst-Pack scheme 34 with a grid with a maximal interval of 0.04Å À1 .

Electrochemical measurements
Electrodes were prepared by pre-mixing Li 2 V x TM 1Àx O 2 F (TM ¼ Ti, Fe, x ¼ 1 or 0.5) with carbon black (acetylene black, Alfa Aesar) in the ball mill to form a composite (300 rpm for 3 h). The obtained composite was mixed with polyvinylidenediuoride binder (PVDF) (Solvay 6050) and N-methyl-2-pyrrolidone (NMP, Alfa Aesar, 99.5%) solution to obtain a slurry with a weight ratio of 70/20/10. The slurry was coated on aluminum foil acting as the current collector and subsequently dried under vacuum at stepwise increasing temperatures up to a maximum of 120 C for 12 h. Aerwards electrodes of 12 mm diameter were punched out. The active material mass loading ranges from 1.4 to 1.9 mg cm À2 with a dry lm thickness between 12 and 21 mm.
For the electrochemical measurements, 2-electrode Swagelok-type cells were assembled using a lithium metal counter electrode, a Li 2 VO 2 F working electrode, 200 mL LP30electrolyte (1 M LiPF 6 in an ethylene carbonate (EC)/dimethyl carbonate (DMC) mixture (1 : 1 by volume, Sigma Aldrich)) and two Whatman glass ber separators. These Li half-cells were assembled in a glovebox under an argon atmosphere. Galvanostatic charge-discharge tests were conducted in the 2electrode setup with an ARBIN BT2000 battery testing system. All electrochemical cycling tests were carried out with a C/5-rate in a voltage range of 1.3 V to 4.1 V vs. Li/Li + . Cycling was performed at room temperature.

HAXPES analysis
For HAXPES analysis, pouch-type cells were prepared in an argon-lled glovebox (O 2 < 2 ppm, H 2 O < 1 ppm). 13 mm diameter electrodes of Li 2 VO 2 F, Li 2 V 0.5 Fe 0.5 O 2 F or Li 2 V 0.5 Ti 0.5 O 2 F were used as working electrodes with a lithium metal (125 mm thick, Cyprus Foote Material) counter electrode and Solupor separator (20 mm thick) soaked in 50 mL LP30-electrolyte. Galvanostatic chargedischarge was performed using a Digatron BTS 600 galvanostat under the same conditions as describe above.
HAXPES was performed on pristine, fully charged and fully discharged electrodes cycled 5 or 50 times. These are abbreviated P, Ch5, DCh5, Ch50, and DCh50 respectively. HAXPES samples were prepared by disassembling the pouch cells, rinsing the working electrode with DMC to remove salt residues, and mounting a piece of the electrode on a sample holder using conductive Cu or carbon tape. All sample transfers were made under an Ar atmosphere. For Li 2 VO 2 F and Li 2 V 0.5 Fe 0.5 O 2 F, measurements were performed at the KMC-1 beamline at BESSY II, Germany 35 using photon energies of 2005 eV. X-rays were monochromatized using a Si(111) double-crystal monochromator and a Gammadata Scienta R-4000 hemispherical analyzer was used to record the photoelectron spectra. Li 2 V 0.5 Ti 0.5 O 2 F samples were measured at the I09 beamline at Diamond light source, UK 36 using photon energies of 2350 eV. X-rays were monochromatized using a Si(111) double-crystal monochromator and a Scienta EW4000 high-voltage electron analyzer was used to record the photoelectron spectra. The probing depth was estimated to be approximately 10 nm using three times the inelastic mean free path (IMFP) for Li 2 VO 2 F derived from the NIST database 37 using material parameters from The Materials Project. 38 HAXPES data analysis and curve-tting were performed using Igor Pro 6.37 soware. All spectra were intensity normalized to the highest intensity peak and binding energy calibrated by shiing the carbon black peak to 284.4 eV. To curve t the transition metal spectra, the spin orbit splitting of the 2p 3/2 and 2p 1/2 doublet peaks was locked to established values (7.3 eV for V, 5.7 eV for Ti, and 13.6 for Fe) while the absolute binding energy values were allowed to vary. In the case of vanadium, only one spin orbit split doublet peak was used to t the spectra, and its binding energy was used to determine the oxidation state. For V 2p and Ti 2p peaks, parameters from the work by Biesinger et al. 39 were used for the tting. The Fe 2p peaks partially overlap with the plasmon of the F 1s peak. 40 By comparing the survey measurements of Li 2 V 0.5 Fe 0.5 O 2 F and Li 2 VO 2 F (ESI, Fig. S11 †) it was determined that the Fe 2p 1/2 peak lies above the uorine plasmon and can thus be used to evaluate iron. The Fe 2p 1/2 was tted using established peak parameters. 41,42 Further details on the tting can be found in the ESI. † To evaluate the composition of the samples, peak areas are calculated and normalized by the photoionization cross section. Ratios between different elements are calculated according to eqn (1): where A i,j is the peak area of elements i and j and s i,j is the theoretical absorption cross section of the corresponding elements as calculated according to the work by Scoeld. 43

Results
We synthesized the target compounds Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F by a high energy ball milling approach, similar to the reported synthesis of Li 2 VO 2 F. 5 High energy ball milling is oen used to synthesize metastable phases or to stabilize phases that normally exist at higher temperatures or pressures. 44 The synthesis procedures reported in the literature so far synthesize the DRS compounds in a one-step process, which in some cases leads to impurities of unreacted precursor compounds. 5,15 In the case of Li 2 VO 2 F, a total ball milling time of 40 hours is suggested. 5 Here, we introduce a two-step approach. In the rst step, we synthesized a LiTMO 2 species by using Li 2 O and the corresponding TM 2 O 3 oxide already exhibits the DRS structure. 45 In a second step, LiF was introduced into the structure as a solid solution by ball milling. This approach has the advantage of signicantly reducing the transition metal oxide precursor impurities in the product. Additionally, we changed the ball mill jar and ball material from tungsten carbide (WC) to less abrasive silicon nitride (Si 3 N 4 ), which reduces the amount of impurities in the synthesized compound. Following this approach, we successfully synthesized two new Li-rich DRS materials, Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F, and compared them with the original Li 2 VO 2 F compound. Rietveld renements based on the synchrotron PXRD of Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F, both mixed with carbon black, are presented in Fig. 1. The broad diffraction peaks of both compounds indicate the nanocrystalline nature of the material, as known for Li 2 VO 2 F (ESI, Fig. S2 †) and similar ball milled materials. 5,9,15,21,45 Rietveld renements are based on the DRS phase (Fm 3m) for both Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F and yield in good ts. The Rietveld renements show that the synthesized products have high purity (the products contain below 1 wt% of Si 3 N 4 impurity). No transition metal precursors (TM 2 O 3 ) were found. The renements yield lattice parameters of a ¼ 4.1342(6)Å for Li 2 V 0.5 Ti 0.5 O 2 F and a ¼ 4.1388(6)Å for Li 2 V 0.5 Fe 0.5 O 2 F, which are slightly larger compared to that of the unsubstituted compound Li 2 VO 2 F (a ¼ 4.1169(4)Å) due to the larger ionic radii of Ti 3+ and Fe 3+ . 46 The rened occupancies of the transition metal positions differ only marginally from the ideal stoichiometry for the Ti and Fe containing phases. Detailed structural parameters obtained from the renement are given in ESI Table S1. † The crystallite size (between 12 and 14 nm) and strain (below 0.6%) were determined from Williamson-Hall plots (ESI, Fig. S3 †). 47 The particle size and morphology of Li 2 V 0.5 Ti 0.5 O 2 F, Li 2 V 0.5 Fe 0.5 O 2 F and Li 2 VO 2 F were investigated by TEM. The results for Li 2 V 0.5 Ti 0.5 O 2 F are presented in Fig. 2. The results for Li 2 V 0.5 Fe 0.5 O 2 F and Li 2 VO 2 F are given in the ESI (ESI, Fig. S4-S6 †). The sample is composed of agglomerated nanocrystalline particles (Fig. 2a). Similar microstructures have been found for Li 2 V 0.5 Fe 0.5 O 2 F and Li 2 VO 2 F. The lattice d-values obtained by selected area electron diffraction (SAED) (Fig. 2b) correspond to the metrics of the DRS structure (Fm 3m) and conrm the results of the Rietveld renement. The highresolution scanning transmission electron microscopy annular dark eld image (STEM-ADF) in Fig. 2c reveals several nanocrystalline (5-10 nm) and some amorphous domains. The corresponding fast Fourier transformation (FFT) of one nanocrystallite is shown in Fig. 2d, matching the orientation along the [101] zone axis of the disordered structure. Despite the local non-uniform mass-thickness contrast in the image, it is possible to identify some differences in the intensity of the atomic columns, indicating variations of the transition metal atomic content. Similar results have been obtained for Li 2 VO 2 F and Li 2 V 0.5 Fe 0.5 O 2 F. Energy dispersive X-ray spectroscopy mapping (EDX) of the materials shows a uniform distribution of elements. However, in the case of Li 2 V 0.5 Fe 0.5 O 2 F the EDX map reveals a small fraction of V-enriched areas of 40-80 nm size, which are not present in the X-ray diffraction pattern (ESI, Fig. S6 †).
To shed light on the structural properties of Li 2 VO 2 F, Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F we investigated the relative structural stability of DRS oxyuoride phases using DFT calculations by comparing the energies of Special Quasi-random Structures (SQS) and ordered prototype structures. These are derived from known ordered oxide structures such as a-NaFeO 2 and g-LiFeO 2 , which are known to be the ground state structure of many lithium transition metal oxides (ESI, Fig. S7 and S8 †). 48,49 The relative structural stability of DRS oxyuoride phases is determined via the energy difference between the SQS and the most stable ordered structure, DE, dened as where E SQS and min(E ordered ) are the total energies of SQS and the most stable ordered structure in meV per atom, respectively. A positive (negative) value of DE indicates that the disordered phase is energetically less (more) stable compared to an ordered phase.
The ordered phase is expected to be more structurally stable than the disordered phase for all considered compounds, because the DRS oxyuorides are in the metastable phase achieved using a mechanochemical ball milling procedure. Furthermore, the decomposition of Li 2 VO 2 F into LiVO 2 and LiF upon heating 50 indicates that the considered compounds may be metastable in general, irrespective of the ordered or disordered phase. Such metastability of the compounds makes it difficult to investigate their relative structural stabilities. However, the ordering propensity, the extent to which the ordered phase is preferred compared to the disordered phase, can be used to assess the relative stability of the compared DRS   In addition to the relative stabilities of the disordered and ordered phases, DFT calculations reveal that the disorder leads to a distribution of oxidation states of the transition metal ions in the compounds. The oxidation states of TM in Li 2 TMO 2 F and TM1 and TM2 in Li 2 TM1 0.5 TM2 0.5 O 2 F (TM, TM1 and TM2 ¼ V, Ti, Fe, respectively) are always 3+ for all of the ordered structures. The oxidation states of V ions are distributed between 2+, 3+ and 4+ in the SQS of Li 2 TMO 2 F (the distribution of oxidation states of the transition metals of the SQS is shown in Table S3 of ESI †). Furthermore, it is observed that the substitution of V with Ti leads to a downward shi in the oxidation state distribution of V ions (between 2+ and 3+) while Ti ions have oxidation states of 3+ and 4+. The opposite happens when V ions are substituted with Fe ions; oxidation states of V ions are distributed between 3+ and 4+ while they are distributed between 2+ and 3+ for Fe ions. A constant value of oxidation states in the ordered phase and its distribution pattern in the disordered phase can be used to determine the extent to which the material is disordered, albeit to a rst order approximation.
The electrochemical performance of the new DRS oxy-uoride compounds Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F was investigated. The proposed theoretical capacity of Li 2 VO 2 F is 462 mA h g À1 based on a 2 e À redox process of the V 3+ /V 5+ couple. Li 2 V 0.5 Ti 0.5 O 2 F has a theoretical capacity of 350 mA h g À1 based on a 1.5 e À redox process assuming additional redox activity of Ti 3+ /Ti 4+ in the low voltage range between 1.5 and 2.0 V. 51 Li 2 V 0.5 Fe 0.5 O 2 F has a theoretical capacity of 226 mA h g À1 assuming electrochemical inactivity of Fe 3+ (2 e À redox process of 50% V 3+ /V 5+ ). Galvanostatic charge-discharge tests of Li 2 VO 2 F, Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F half cells have been conducted. The cycling performance is shown in Fig. 3. The materials have been cycled within a potential range of 1.3 and 4.1 V vs. Li/Li + with a C/5-rate. Li 2 VO 2 F shows the highest rst discharge capacity of all three compounds (Fig. 3a) of around 330 mA h g À1 , which is in good agreement with the literature, accompanied by rapid capacity fading known from previous reports. 5,14 45% of the initial discharge capacity is lost aer 25 cycles. Aer 50 cycles the discharge capacity is already below 150 mA h g À1 , which corresponds to less than 40% capacity retention (Fig. 3b). Both substituted compounds, Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F, exhibit a lower discharge capacity of 285 mA h g À1 and 218 mA h g À1 in the rst cycle, respectively. The discharge capacity of Li 2 V 0.5 Ti 0.5 O 2 F is 67 mA h g À1 higher compared to Li 2 V 0.5 Fe 0.5 O 2 F, which may be explained by additional contribution to the capacity of the Ti 3+ /Ti 4+ redox couple. The capacity fading is signicantly reduced for both substituted compounds; Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F retain 81% and 83% of the initial discharge capacity aer 25 cycles and 66% and 73% aer 50 cycles, respectively. The coulombic efficiency (Fig. 3a) is improved for both new materials (around 97% for 50 cycles) compared to Li 2 VO 2 F (around 93%). Altogether, the substitution of V with 50% Ti or Fe clearly improves the cycling performance compared to Li 2 VO 2 F. Li 2 V 0.5 Fe 0.5 O 2 F shows the best cycling stability over 50 cycles, whereas Li 2 V 0.5 Ti 0.5 O 2 F exhibits the highest overall discharge capacities. The corresponding voltage proles of Li 2 VO 2 F, Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F exhibit a steep and sloping prole shape enhanced by the disorder suggesting a single-phase insertion process for Li + (ESI, Fig. S9 †). 5,6,14 No voltage plateaus are observed. The average discharge voltage of Li 2 VO 2 F is about 2.53 V with a voltage To understand the redox processes occurring during electrochemical cycling the differential capacity dQ/dV plots are shown in Fig. 4. In Li 2 VO 2 F, the oxidation of V 3+ to V 4+ is located in the area of 2.6 V and that of V 4+ to V 5+ is located above 3.5 V (indicated with dashed lines). 14,15 For Li 2 V 0.5 Ti 0.5 O 2 F, the assumed redox peaks of vanadium (dashed lines) are slightly shied to higher voltages and exhibit the highest overpotentials, which may be related to kinetic effects. Furthermore, additional peaks in the charge and discharge directions are observed at 2.2 V and 1.8 V, respectively (dotted lines). These peaks are not present in the samples that do not contain Ti (Li 2 VO 2 F and Li 2 V 0.5 Fe 0.5 O 2 F) and thus are expected to originate from the Ti 3+ /Ti 4+ redox couple leading to an additional discharge capacity. Li 2 V 0.5 Fe 0.5 O 2 F shows the smallest overpotential and behaves like Li 2 VO 2 F in the low voltage regime indicating a similar redox behavior, but differs in the voltage region above 3.5 V during charging. This deviation may be related to processes at high voltages associated with irreversible reactions affecting the cycling stability. Upon extended cycling the dQ/dV plot of Li 2 VO 2 F tends to a attening differential capacity peak response, indicating a loss of V-redox activity. 20 In contrast, Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F preserve the characteristic redox peaks for a longer cycling period. This suggests that the maintained electrochemical activity of the TM is related to the improved cycling stability of the materials. Like for Li 2 VO 2 F, the ex situ PXRD pattern of Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F aer extended cycling does not exhibit any development of new crystalline phases (ESI, Fig. S10 †). 5 To further analyze the materials' surface stability, HAXPES was used to probe the redox activity of the transition metals and the surface layer evolution. Since HAXPES is a surface sensitive technique it has commonly been used to study the surface layers built up on the active material, known to be crucial for the cycling performance. 52,53 Thus, to understand the improved capacity retention of the substituted materials the O 1s, V 2p and C 1s spectra are analyzed for pristine (P) and samples cycled 5 or 50 times in both charged (Ch5 and Ch50, respectively) and discharged (DCh5 and DCh50, respectively) states. The photon energy for the measurements is chosen so that both the outer layers of the active material and the surface layer can be probed.
The O 1s and V 2p spectra are shown in Fig. 5. Five different peaks are used to t the data, from le to right corresponding to carboxyl/hydroxyl compounds ($534 eV), carbonates ($532 eV), metal oxide (530 eV) and vanadium that is detected with a spin orbit splitting of 7.33 eV at $517 and 524 eV according to peak parameters summarized by Biesinger et al. 39 The energy difference of the O 1s metal oxide peak and V 2p 3/2 can be used to determine the oxidation state of V, where a larger value corresponds to a lower oxidation state. The values obtained from the tting are presented in Table 1  and V 5+ in the region probed. The deviation from an average oxidation state of 3+ might be related to surface oxidation, as seen also for other vanadium oxides. 39 The mix of oxidation states is seen from the rather large FWHM of the V 2p peaks. This can be expected due to the disordered structure, where vanadium can be found with a different coordination of oxygen and uorine, affecting the binding energy. Upon h charge all samples are as expected close to a fully oxidized V 5+ state. Aer the following h discharge Li 2 V 0.5 Fe 0.5 O 2 F and Li 2 V 0.5 Ti 0.5 O 2 F return close to their respective pristine state, while the vanadium in Li 2 VO 2 F is no longer redox active and stays in a highly oxidized state. For the Li 2 VO 2 F material this reduced redox activity of vanadium has previously been suggested to be linked to a partial oxidation of oxygen forming reactive compounds that leads to a breakdown of the active material, starting at the surface. 20 In this context both the Li 2 V 0.5 Fe 0.5 O 2 F and Li 2 V 0.5 Ti 0.5 O 2 F materials clearly show improved reversibility of the vanadium redox behavior. Still, aer 50 cycles the materials show less to no redox activity of vanadium in the depth region probed by HAXPES. This trend can also be followed in the FWHM, which changes upon cycling. Especially on h discharge a broadening can be seen, indicating that some of the material can no longer return to its original state. This broadening is most signicant for the unsubstituted material. The smaller FWHM aer 50 cycles can be explained by a more uniform V 5+ state of the probed material, in combination with that V 5+ exhibits narrower peaks than V 4+ and V 3+ , since V 5+ does not have any unpaired valence electrons. 39 To gain a deeper insight into the improved redox activity of the substituted materials it is interesting to look at the intensity ratio between vanadium and the metal oxide (MO) peak. 20 In Table 2 the V : MO ratios are presented, with the oxygen content normalized to two for easy comparison. According to the structural formulas, the V content should be 1 for Li 2 VO 2 F and 0.5 for Li 2 V 0.5 Fe 0.5 O 2 F and Li 2 V 0.5 Ti 0.5 O 2 F. For the pristine materials the ratio is slightly higher than expected for the Li 2 VO 2 F material, while for the Li 2 V 0.5 Fe 0.5 O 2 F and Li 2 V 0.5 Ti 0.5 O 2 F materials the ratio is slightly lower. Comparing charged and discharged samples, it is seen that the relative ratio is higher in the discharged samples for Li 2 VO 2 F and Li 2 V 0.5 Fe 0.5 O 2 F, while the Li 2 V 0.5 Ti 0.5 O 2 F material shows a small but opposite trend. For Li 2 VO 2 F the V : MO ratio increases signicantly during cycling together with a binding energy shi of the MO peak. As discussed in detail in another study, 20 the relative increase and decrease of vanadium compared to oxygen can be coupled both to possible oxygen redox processes as well as the formation and dissolution of a surface layer containing vanadium. The same trend cannot be seen for the Li 2 V 0.5 Fe 0.5 O 2 F and Li 2 V 0.5 Ti 0.5 O 2 F materials. Only a slight increase of the V : MO ratio is seen for Li 2 V 0.5 Fe 0.5 O 2 F, while for Li 2 V 0.5 Ti 0.5 O 2 F a close to constant ratio is obtained up to 50 cycles. Additionally, only small binding energy shis of the MO peak are seen (<0.2 eV). These results clearly indicate that substitution with iron and titanium mitigates the detrimental reactions causing vanadium dissolution and incorporation in the surface layer and thus improves the chemical stability of the materials.
Looking further at the other transition metals (ESI, Fig. S12 †), both iron and titanium are found to be partially redox active at the surface. Iron is found in a mix of Fe 3+ and Fe 2+ in the pristine material and upon cycling the Fe 3+ content increases aer h charge while more Fe 2+ is found aer subsequent discharge. Titanium is predominately found in the Ti 4+ state for the pristine material, with some amount of Ti 3+ upon discharge. As already mentioned for vanadium, the deviation from the 3+ oxidation state indicates that an oxidized surface layer is present already aer the synthesis of the materials. This kind of passivating surface lm is oen seen on cathode materials. 54,55 The surface layer evolution during cycling is evaluated using the carbon spectra, as shown in Fig. 6. Here it is particularly interesting to look at the relative intensities between the carbon black (CB) bulk peak (shaded in red) and the hydrocarbon (C-H) surface peak. A relatively lower CB peak intensity signies a thicker surface layer. Starting with the unsubstituted material (Fig. 6a), a buildup of a surface layer is seen upon charge, followed by its partial dissolution upon discharge. This is consistent with previous results for Li 2 VO 2 F. 20 The substituted samples on the other hand show a stabilized surface aer 50 cycles (similar Ch50/DCh50 spectra). Especially Li 2 V 0.5 Ti 0.5 O 2 F (Fig. 6b) shows a rather thin and stable surface layer with similar spectra for all samples. For Li 2 V 0.5 Fe 0.5 O 2 F the surface layer is of similar thickness compared to the Li 2 VO 2 F sample, but the layer is more stable and no dissolution is observed aer 50 cycles.  The other peaks in the C 1s spectra stem from the PVDF binder (two peaks at $286 and $290 eV) and different carbon oxygen compounds (at $286.5, 288 and 290 eV), typically stemming from electrolyte degradation. The surface layers are seen to consist of mostly hydrocarbons and some C-O compounds. On Li 2 V 0.5 Fe 0.5 O 2 F (Fig. 6c) the surface layer is built up with a relatively larger amount of C-O compounds, probably stemming from electrolyte breakdown. The presence of a surface layer indicates that some side reactions occur for all materials, but a more stable layer, as found on the substituted materials, can limit the extent of these reactions by passivating the surface.

Discussion
Combining the results of the synthesis, the DFT calculations, the electrochemical experiments and the HAXPES analysis, the results can be discussed related to disorder of the structure and related to the electrochemical cycling behavior of the materials. The disorder in the crystal structure of the pristine TM-oxyuorides leads to a distribution of oxidation states deviating from the original 3+ state in the ordered structures for the presented compounds as has been revealed by the DFT calculations. The computed mixed 2+, 3+, and 4+ oxidation states for vanadium in Li 2 VO 2 F are consistent with the HAXPES results, considering additional contribution of Li-deciency and oxidation at the surface. 14,20 The DFT and HAXPES results further agree with the mixed 2+ and 3+ oxidation states of iron in Li 2 V 0.5 Fe 0.5 O 2 F, which lead to a shi for vanadium to 4+ and 5+. For Li 2 V 0.5 Ti 0.5 O 2 F, where the titanium oxidation state is 3+ and 4+, the substitution instead lowers the oxidation state of vanadium to 3+. In addition, the DFT calculations showed that the metastable disordered rock salt structure was stabilized when partially substituting vanadium with iron or titanium.
The electrochemical cycling behavior of the substituted materials differs from that of Li 2 VO 2 F. Both substituted materials exhibit a signicantly more stable cycling behavior but a lower discharge capacity. The differential capacity analysis reveals additional capacity contributions of Li 2 V 0.5 Ti 0.5 O 2 F compared to Li 2 V 0.5 Fe 0.5 O 2 F and Li 2 VO 2 F indicating redox activity of titanium in the bulk. Whittingham et al. determined the oxidation state of vanadium in Li 2 VO 2 F in the charged state by X-ray absorption spectroscopy. They observed an average oxidation state of only 4.2+ when charged to 4.1 V vs. Li/Li + . 14 Complete oxidation to V 5+ could not be achieved. The HAXPES results of Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F conrm a partial redox activity of iron and titanium at the surface (ESI, Fig. S12 †). Furthermore, we believe that some of the additional capacity of Li 2 V 0.5 Ti 0.5 O 2 F compared to Li 2 V 0.5 Fe 0.5 O 2 F stems from the fact that substitution with titanium promotes a complete use of the V 3+ /V 5+ redox couple. Further analysis by X-ray absorption techniques may shed light on the different redox reactions occurring in the bulk material.
In recent reports, DRS materials are cycled to high potentials up to 4.8 V vs. Li/Li + , which facilitates anionic redox activity of oxygen in the lattice that leads to additional capacity contribution. 11,[14][15][16]21 Materials containing Ti, Li 1.2 Ti 0.4 Mn 0.4 O 2 for instance, experimentally sustain a stable oxygen-redox reaction above 4.1 vs. Li/Li + . 10 The observed shi in binding energy of the MO peak together with the relative changes in intensity between the MO peak and vanadium indicates that such anionic redox activity occurs in the surface region of the Li 2 VO 2 F material already when cycling to 4.1 V vs. Li/Li + . 56 In Li 2 VO 2 F this is believed to create highly reactive oxygen atoms in the lattice leading to the instability of the surface, as discussed in more detail in the work of Källquist et al. 20 The reaction between the oxidized lattice oxygen and the electrolyte creates an interfacial layer rich in vanadium in oxidation state 5+. Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F show a higher reversibility of the V redox reaction according to the differential capacity analysis and the HAXPES results. In the dQ/dV plot, the substituted materials show a reduced irreversible capacity contribution above 3.5 V up to. 4.1 V vs. Li/Li + . At the same time, the HAXPES data only show small changes of the V : MO-ratio and the binding energy of the MO peak aer 50 cycles for Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F. Together this suggests a reduced reactivity of the lattice oxygen species at the surface in these new materials when cycled to 4.1 V vs. Li/Li + . We propose this as an explanation for the improved cycling stability. This is further supported by the cycling performance when the materials are cycled up to 4.5 V vs. Li/Li + . These results (see ESI, Fig. S13 †) show additional contribution to the capacity for all three compounds, possibly originating from anionic redox activity, but are also accompanied by a reduced cycling stability.
Additional contribution to the cycling stability likely comes from mitigation of the dissolution and rebuilding of the surface layer for the substituted materials that otherwise is seen on Li 2 VO 2 F. Although the thickness of the surface layers on Li 2 V 0.5 Ti 0.5 O 2 F and Li 2 V 0.5 Fe 0.5 O 2 F varies, both are to a large part preserved during cycling. This is in agreement with a recent