Controlled Hydroxy-Fluorination Reaction of Anatase to Promote Mg 2+ Mobility in Rechargeable Magnesium Batteries

In anatase TiO2, substituting oxide anions with singly charged (F,OH) anions allows the controlled formation of cation vacancies, which act as reversible intercalation sites for Mg2+. We show that ion-transport (diffusion coefficients) and intercalation (reversible capacity) properties are controlled by two critical parameters: the vacancy concentration and the local anionic environment. Our results emphasise the complexity of this behaviour, and highlight the potential benefits of chemically controlling cationic-defects in electrode materials for rechargeable multivalent-ion batteries.


Synthesis
Anatase featuring different compositions/vacancy concentrations have been prepared following a previously reported method. 1,2 Briefly, 1.2 mL aqueous hydrofluoric acid solution was added to the mixture of 24.8 mL isopropanol and 4 mL titanium isopropoxide in the Teflon liner cup within a stainless-steel autoclave. After sealing the autoclave, the mixed solution was heated inside an oven at 90 °C and 130 °C for 12 h, respectively. After cooling down to room temperature, the obtained white precipitates were washed with ethanol and centrifuged, then dried at 100 °C under air for 10 h. The recovered samples were further outgassed at 150 °C overnight under primary vacuum prior to physico-chemical and electrochemical characterizations. X-ray diffraction patterns ( Figure S1) confirmed the phase purity. To determine the chemical composition, we assessed the vacancy content by structural analysis of total scattering data, evaluated the fluorine content using solid-state 19 F NMR and the OH content was deduced according to the general chemical formula Ti1-x-y¨x+yO2-4(x+y)F4x(OH)4y. 3 Figure S1. X-ray diffraction patterns of the samples prepared at 90 and 130 °C.
Chemical magnesiation was performed according to the previously reported procedure. 4 Briefly, ethylmagnesium bromide solution (3.0 M in diethyl ether, Sigma-Aldrich) was added drop-by-drop to a solution containing diethyl ether and the solid host Ti0.78¨0.22O1.12F0.4(OH)0.48. The mixture was stirred for 48 hours at room temperature. The obtained powder was collected by filtration, washed with anhydrous diethyl ether solvent, and dried under vacuum. All subsequent operations were carried out in an argon-filled glove box.

Electrochemistry
Swagelok-type three-electrode cell was used for the electrochemical characterization. The working electrode was composed of 80 wt% active material, 10 wt% conductive carbon (SuperP, Timcal), and 10 wt% polyvinylidene difluoride (PVDF, Aldrich). The suspension of those chemicals in N-methyl-2-pyrrolidone (NMP, Sigma-Aldrich) was hand milled using a mortar, and then drop coated on a Mo foil (99.95%, Alfa Aeser) at the geometrical active mass density of 2 mg cm -2 . Mg metal plate (99.9%, Good fellow) was used as the reference electrode and the counter electrode. 0.2 mol L -1 2PhMgCl -AlCl3 / THF was prepared by dropping 0.5 mol L -1 AlCl3 / THF (Sigma-Aldrich) into 2 mol L -1 PhMgCl / THF (Sigma-Aldrich) under agitation for 12 hours, and it was used as the electrolyte, and borosilicate glass-fiber filter paper (Whatmann grade GF/A) was used as the separator. The measurements were carried out at 25 o C. Galvanostatic dischargecharge measurements were performed at different current densities in the potential range of 0.05-2.3 V vs. Mg 2+ /Mg, and the specific capacities were calculated based on the mass of the active material on the electrode.

Galvanostatic Intermittent Titration Technique
Galvanostatic intermittent titration (GITT) 5-7 experiments were performed for the first discharge ( Figure S2). In these experiments, a current density of 10 mA g -1 was applied for 1h followed by a 10h relaxation period from the open circuit voltage to 0.05 V. Diffusion coefficient was estimated according to the following equation: Here, F is Faraday's constant, z is charge number of Mg (z = 2), S is the surface area of electrode, I is the current, VM is the molar volume of the host lattice, and L is characteristic length of electrode materials.  reveal the disappearance of the line characteristic of Ti IV o2-F species (from or even before 0.07 Mg 2+ per formula unit) and a strong intensity decrease for the resonance assigned to Ti IV 2o-F species. This evidences a filling of the vacancies by Mg 2+ , preferentially and firstly in di-vacancy systems (becoming single-vacancies). Based on the 19   The asterisks indicate the main spinning sidebands. The dashed lines indicate the 19 4,10 Tentative assignment of the NMR resonances to the various species which are supposed to occur in the chemically magnesiated samples and fits of the spectra (a) and (c) have been reported previously. 4 The fit of the spectrum (b) is presented in Figure S4.  Table S2).   Table 4. Estimated proportions of anionic environments in the vicinity of titanium (TiX6) and vacancy (¨X6) assuming random distributions of the anions (X = F, O(OH)) in Ti1-x-y¨x+yO2-4(x+y)F4x(OH)4y samples.

Density Functional Theory (DFT) calculations
Our density functional theory (DFT) calculations were performed using the VASP, 12,13 with valence electrons described by a plane-wave basis with a cutoff of 500 eV.
Interactions between core and valence electrons were described using the projector augmented wave (PAW) method, 14  giving cell stoichiometries of Ti127O252X4. These calculations agree with our previous study, that found that fluoride ions preferentially occupy sites adjacent to the titanium vacancy in equatorially-coordinated sites. 1 We find that OH shows the same preference for occupying anion sites adjacent to the titanium vacany. 21 Magensium intercalation into Ti1-x-y¨x+yO2-4(x+y)F4x(OH)4y was modelled using 4 × 4 × 2 supercells, with 1 Ti vacancy, and 4 charge compensating XO species (X=F, OH) occupying four anion sites adjacent to the vacancy. We calculated intercalation energiesfor 4X=(4F, 3F+OH, 2F+2OH, F+3OH, and 4OH). In the case of 4X=2F+2OH, we considered like anions arranged in an opposite (trans) equatorial site pair configuration. We have previously shown that this gives a similar intercalation energy as for the adjacent (cis) equatorial site pair configuration. 21 Calculations of the intercalation energy of Mg at vacancyadjacent sites were performed using the same 4 × 4 × 2 cell. For each anion combination considered, we placed a Mg ion at each of eight adjacent interstitial sites and performed a geometry optimization keeping the cell shape and volume fixed. Individual calculations were deemed optimised when all atomic forces were smaller than 0.01 eV Å -1 . All calculations were spin polarized, and used a 4 × 4 × 2 Monkhorst-Pack grid for sampling k-space in the single unit cell, and only the gamma-point for the 4 × 4 × 2 cells. To calculate intercalation energies, reference calculations for metallic Mg was performed using the same convergence criteria as above. We considered a 2-atom cell for Mg, with a 16 × 16 × 16 Monkhorst-Pack grid for k-space sampling. Data sets containing these DFT calculation inputs and outputs are available at the University of Bath Data Archive 20,22 . Code to produce figure 3b from these data is included in dataset Ref 22. Analysis scripts containing calculations of competing anion configurations are available as part of an open-source repository as reference, 21 published under the MIT license.