Surface Engineering Strategy Using Urea To Improve the Rate Performance of Na2Ti3O7 in Na‐Ion Batteries

Abstract Na2Ti3O7 (NTO) is considered a promising anode material for Na‐ion batteries due to its layered structure with an open framework and low and safe average operating voltage of 0.3 V vs. Na+/Na. However, its poor electronic conductivity needs to be addressed to make this material attractive for practical applications among other anode choices. Here, we report a safe, controllable and affordable method using urea that significantly improves the rate performance of NTO by producing surface defects such as oxygen vacancies and hydroxyl groups, and the secondary phase Na2Ti6O13. The enhanced electrochemical performance agrees with the higher Na+ ion diffusion coefficient, higher charge carrier density and reduced bandgap observed in these samples, without the need of nanosizing and/or complex synthetic strategies. A comprehensive study using a combination of diffraction, microscopic, spectroscopic and electrochemical techniques supported by computational studies based on DFT calculations, was carried out to understand the effects of this treatment on the surface, chemistry and electronic and charge storage properties of NTO. This study underscores the benefits of using urea as a strategy for enhancing the charge storage properties of NTO and thus, unfolding the potential of this material in practical energy storage applications.


Introduction
The increasing demand for electrochemical energy storage devices has resulted in rapid development and utilisation of Liion batteries (LIBs) in recent years. [1] LIBs are widely used in portable electronic devices, electric/hybrid vehicles and smart grid systems. [1] However,t he scarcityo fl ithium sourcesc ombined with their growingd emandh ave motivated the development of alternative storage technologies. [2] Sodium-ion batteries (SIBs) offer ap romising low-cost energy storage alternative to LIBs owing to the abundance of sodium sourceso n Earth. [3] Furthermore, inexpensive aluminiumc urrent collectors can be used on the anode side, instead of the more expensive copperi nL IBs. [4] Nevertheless, one of the greatest challenges relatedt ot he full incorporation of SIBs into the market is finding suitable and safe anode materials that deliver high and stable capacities at low voltages in classic organic electrolytes. [4] Titanium-based materials arise as one of the most promising candidates due to their non-toxicity,l arge abundance and low manufacturing cost. [5] Among thisf amily of compounds, there has been ag rowing interest in exploring the zig-zag layered Na 2 Ti 3 O 7 (NTO) phase as aS IB anode material in the last decade. [6] NTO was first tested as an anode in SIBs about a decadea go, where it was shown to reversibly intercalate up to 2N ai ons per formula, resulting in ah ight heoretical capacity (177 mAh g À1 )a tal ow and safe average potentialo f0 .3 Vv s. Na + /Na. [6] The insertiono fN a + ions in NTO is commonly accepted to proceed throughatwo-phase reaction mechanism that leads to the formation of the end-discharge product Na 4 Ti 3 O 7 (with the correspondingr eductiono f 2/3 Ti 4 + ions to Ti 3 + ). [6,7,8] More recent studies have shownastable and partially sodiated intermediate phase with composition Na 3Àx Ti 3 O 7 which forms upon the first discharge process. [7] Unfortunately, the prospects for practical application of NTO in SIBs are compromised by sluggish Na + ion diffusion kinetics due to its structural distortion upon uptake of Na + ions andt he intrinsically electronicallyi nsulating nature of NTO, associated with a large bandgap of 3.7-3.9 eV. [3,9] These result in poor electrochemicalp erformance at high charge/discharge rates, limiting the use of NTO in high powerapplications. [3,10] Several research strategies including carbon-composite fabrication, [7,8,11] nanostructuring, [12][13][14][15][16] doping [17,18] and surface defecte ngineering [9,10,19,20] have shown to improvet he electrical/ionic conductivity and electrochemical performance of NTO. Surface defect engineering typically involves the introduction of oxygen vacancies in the NTO structure, whicha ct as n-type defects, improvingc harge transferp rocesses by narrowing the bandgap. Typically, synthetic methods that introduce oxygen vacancies in NTO involve the direct use of H 2 (which may raise safety concerns), limitingt heir practical applications. [9,10] Therefore, considering the potential benefits associated to the creation of oxygen-deficient materials, it is desirable to develop as afe, controllable and affordable synthesis methodt oo btain NTO with improved electrochemical performance through an effective surfaceengineering method. In this context, urea (CH 4 N 2 O) has been successfully used as ar educing agentt oc reate oxygen vacancies in metal oxide photocatalystss uch as WO 3 and BiOBr. [21,22] At relativelym ild temperatures, urea decomposes into ammonia which furtherd ecomposes to produce reactive H 2 that removes oxygen from the structure. [21] Herein, we have prepared NTO with different levels of oxygen vacancies by mixing as-synthesised NTO with ureaa td ifferent concentrations(5, 10 and 20 wt.%) beforeannealing in aN 2 atmosphere at 450 8Cf or 2h.T he 20 wt. %u rea sample wasf ound to be the optimal concentration,s howing initial discharge capacities of 316 mAh g À1 (1 C) and 272 mAh g À1 (2 C) when tested as an anode in Na-ion half-cells. After 100 cycles,d ischarge capacitieso f1 54 mAh g À1 (1 C) and1 45 mAh g À1 (2 C) were obtained, which are significantly highert han those observed in the pristine material 106 mAh g À1 (1 C) and 90 mAh g À1 (2 C)). The improved electrochemical performance is attributed to the higherN a + ion diffusion coefficient, higher charge carrier density and reduced bandgapo bserved in the urea-treated sample. As will be described later,t he urea treatment not only produces oxygen vacancies but also leads to the formation of Ti 3 + and TiÀOH species and Na 2 Ti 6 O 13 ,w hich together are responsible for the improved rate performance. Our work will examine the effects of the urea treatment on the surface, chemistry and electronic and charges toragep roperties of NTO, with ac ombination of experimental characterisation techniques and computational studies based on density functional theory (DFT).

Experimental Section
Synthesis of Na 2 Ti 3 O 7 :N TO was synthesised by as olid-state reaction method using as toichiometric mixture of TiO 2 (Fisher Scientific, 98 %) and anhydrous Na 2 CO 3 (Sigma-Aldrich, 99.9 %). These were mixed in ap lanetary ball mill at 400 rpm, followed by ah eat treatment at 800 8Cf or 20 h. The synthesised NTO powders were mixed with different concentrations of urea, CH 4 N 2 O, (Thermoscientific, 99.5 %) (5, 10 and 20 wt %), followed by an annealing treatment at 450 8Cfor 2hunder an N 2 atmosphere. The obtained powders, herein described as 0U, 5U, 10U and 20U, correspond to NTO mixed with 0, 5, 10 and 20 wt. %u rea, respectively.T he as-prepared powders were mixed with 50 wt %s ucrose, C 12 H 22 O 11 , (Sigma-Aldrich, 99.5 %) and then subjected to ap yrolysis treatment at 700 8Cf or 5h under flowing argon to carbon-coat the powders to allow for ag ood comparison between the pristine material and the urea-treated samples while improving the overall electronic conductivity of the samples. [8] Physicochemical characterisation:P owder X-ray diffraction (PXRD) data were recorded at room temperature using aS martlab diffractometer (Rigaku Corporation) equipped with a9kW Cu rotating anode (l = 1.54056 )o perating in reflection mode with Bragg-Brentano geometry.D ata were collected in the 5-70 8 2q range at as can speed of 0.02 8 s À1 .T he NTO structure was refined against powder X-ray diffraction data using the Rietveld method, with the GSAS-EXPGUI software interface. [23,24] The peak shapes were modelled with aG aussian-Lorentzian function and the background, lattice parameters, atomic positions and thermal parameters were refined. The thermal parameters for individual Na, Ti and Oa toms were refined isotropically and constrained to be identical. The occupancy for all atoms was fixed to n = 1.
The microstructure of the samples was examined using af ield emission scanning electron microscope (FESEM;J EOL JSM-7800F) operated at 5kVa nd 5mA. Energy-dispersive X-ray analysis (EDX) was carried out at 20 kV to assess the elemental composition, using the AZtecEnergy software. Before the analysis, powders were coated with au niform layer of Au/Pd by sputtering deposition to provide surface conductivity and prevent surface charging.
Transmission electron microscopy (TEM, JEOL 2200FE) was carried out at 100 keV to assess the morphology and thickness of the carbon coating layer after the pyrolysis treatment. Initially,s amples were prepared by mixing the as-prepared powders with dried acetonitrile using an ultrasonication method. The suspension was then drop cast onto aT EM grid inside the glovebox, dried under vacuum and transferred to the microscope under an Ar atmosphere.
The amount of carbon in the samples was determined by thermogravimetric analysis (TGA;T AI nstruments Q5000IR) in air by heating the powders from ambient temperature to 700 8Cu sing a heating ramp of 10 8Cmin À1 .
Electron paramagnetic resonance (EPR) spectroscopy was performed at room temperature on aB ruker MicroEMX spectrometer equipped with aB ruker super high Qr esonator with am icrowave frequency of 9.87 GHz, microwave power of 1mW, field modulation of 100 kHz and modulation amplitude of 4G .T he field calibration was carried out using 2,2-diphenyl-1-picrylhydrazyl (DPPH) as as tandard and measurements were normalised to the sample weight.
X-ray photoelectron spectroscopy (XPS) was carried out using a PHI 5500 XPS instrument with an Al Ka X-ray source (1486.6 eV). Powder samples were mixed with as mall amount of carbon black, using am ortar and pestle, to provide good electronic conductivity for the measurements. The energy was calibrated to the graphitic carbon (C=C) peak in the C1 ss pectra (284.0 eV) for each sample. Data were analysed with the CasaXPS package software, employing the Gaussian-Lorentzian peak shape GL (30). 23 Na magic-angle spinning (MAS) NMR spectra were acquired using a7 00 MHz Bruker Avance III HD WB spectrometer at am agnetic field of 16.4 T. Experiments were performed using aB ruker 3.2 mm probe at aM AS rate of 10 kHz. Spectra were referenced relative to 1 m NaCl (aq) solution using the 23 Na resonance of solid NaCl at 7.5 ppm as asecondary reference.
UV-visible spectroscopy was performed with aC ary500 spectrometer in the 200-500 nm range using an integrating sphere to acquire only diffusiver eflectance of the electromagnetic radiation.
Electrochemical characterisation:T he electrochemical performance of the sodium titanate samples synthesised in this work was tested using stainless steel CR2032 coin cells, Na metal as the counter/reference electrode (Alfa Aesar Merck), 1 m NaPF 6 (99 % Alfa Aesar) in ethylene carbonate (EC): diethyl carbonate (DEC) solvent (battery grade, Gotion) (1:1 v/v)ast he electrolyte and aWhatman micro glass fibre separator.T he liquid organic electrolyte was dried for several days using activated molecular sieves (0.4 nm pore diameter,M erck) before use. The assembly and electrode preparation were carried out in an Ar filled glovebox (MBraun, H 2 O and O 2 < 0.1 ppm). The electrode preparation involved mixing of the active material (sodium titanate samples) with carbon black (Super P) (99 %A lfa Aesar) and polyvinylidene binder (PVDF Kynar, 99 %A lfa Aesar) in aw eight ratio of 70:20:10, respectively.E lectrode slurries were prepared by adding af ew drops of N-methyl-2pyrrolidone (NMP) (anhydrous, 99 %A lfa Aesar) to the electrode mixture, which was stirred for 12 h. The obtained slurry was coated uniformly onto an aluminium foil using ad octor blade to form a film of 200 mmt hickness which was then dried at 80 8Cu nder vacuum for 12 hi nt he antechamber of the glovebox. Subsequently,t he electrodes were cut into circular disks of 19 mm diameter with aload of active material of ca. 1.25 mg cm À2 .
Galvanostatic charge/discharge measurements were performed on ab attery tester (Neware battery system, current range:1 -10 mA) in the voltage range 0.01-2.5 Vv s. Na + /Na at different rates (0.1, 0.2, 1a nd 2C). Cyclic voltammetry measurements were conducted on an Ivium potentiostat (Alvatek) in the voltage range 0.01-2.5 V vs. Na + /Na at different scan rates (0.05, 0.1, 0.2 and 0.3 mV s À1 ). Mott-Schottky measurements were carried out on an Ivium potentiostat (Alvatek), in the voltage range 0.01-2.5 Vv s. Na + /Na using af requency of 500 Hz with as can step of 50 mV.E lectrochemical impedance spectroscopy (EIS) data were collected on an Ivium potentiostat (Alvatek) with an AC amplitude of 10 mV in the frequency range between 0.05 and 10 5 Hz. Data were acquired during the first discharge process at OCV ( % 2.5 V), 1, 0.4, 0.2 and 0.01 Vv s. Na + /Na. Ab initio calculations:A ll the calculations performed in this work used the density functional theory (DFT) method as implemented in the Vienna Ab initio Simulation Package code. [25,26] The projector augmented wave approach [27] was employed to describe the interaction between the core and valence electrons. The electron configurations Na (3s 1 ), Ti (3d 3 4s 1 ), and O( 2s 2 2p 4 )w ere treated as the valence electrons. Brillouin zones for all compounds were sampled such that the k-points were converged in an accuracy of the total energy in 0.001 eV atom À1 (Table S1) and the plane-wave cut-off was set to be 500 eV to sufficiently converge the total energy to within 0.01 eV atom À1 .I nt his relaxation, the atomic positions, lattice vector,a nd cell angle were allowed to relax. All calculations were deemed to be converged when the forces on all atoms were less than 0.01 eV À1 .
The revised Perdew-Burke-Ernzerhof Generalised Gradient Approximation (GGA) functional (PBEsol) [28] was used for all phase stability calculations including enthalpy and vibrational entropy.P BEsol has accurately reproduced lattice parameters and lattice dynamics in solid systems while maintaining ar elatively low computational cost. [29,30] The enthalpy (H) was calculated for 20 different phases of the Na-Ti-O system obtained from the Materials Project database (Table S1). [31] The vibrational entropy (S vib )w as calculated using the supercell and finite displacement approaches, as implemented in the Phonopy package ( Figure S1). [32a] For the entropy calculations, all the structures were initially relaxed to aforce convergence criterion of 0.0001 eV À1 .T he harmonic force constants and associated vibrational entropy were then calculated by creating atomic displacements in 4 4 4( 128 atoms), 1 2 3( 36 atoms), 3 where T is the absolute temperature. The free energies of formation for Na x Ti y O z (DG <f,Na x Ti y O z )c an then be calculated as [Eq. (2)]: where G Na , G Ti ,a nd G O 2 are the free energies of bulk Na, Ti,a nd O 2 gas.
All electronic structure and band alignment calculations were performed using the screened hybrid functional (HSE06), [33] in which 25 %o fe xact non-local Fock exchange is added to the PBE [34]

Results and Discussion
Structure and Characterisation Figure 1a shows the powder X-ray diffraction (PXRD) data for pristine NTO (0U) and urea-treatedN TO samples (5-20U).F igure S2 showst he PXRD pattern of 0U (black) togetherw ith the calculated (red) and difference( blue) profiles obtained by Rietveld refinement. The resulting structuralp arameters are shown in Table S2. Ag ood agreement is observed between the experimentald ata and the pattern calculated with NTO structural model (space group P12 1 /m1) reportedi nt he literature. [37,38] Furthermore, Figure1bs hows that the most intense diffraction peak corresponding to the (001) crystal plane is present at identical2 q values (i.e. 10.68)i na ll the samples,i ndicating that the interlayer distance (d)does not change after the urea treatment (d % 8.57 ). As econdary monoclinicp hase with composition Na 2 Ti 6 O 13 (space group C2/m,I CSD2 3877)i so bserved in sample 10U and its content increasesp roportionally with the amount of urea used, implying that NTO undergoes partial decomposition during the urea treatment. Reports have shown that given the structural similarities between both sodium titanate phases, NTO may be converted to Na 2 Ti 6 O 13 through sodium and oxygen loss upon heating at high temperature (950 8C) in air [39] or at mild temperatures (400-500 8C) in ar educinga tmosphere. [10,40] Furthermore, some reports have shown the formation of Na 2 Oi na ddition to Na 2 Ti 6 O 13 during thermald ecomposition of NTO. [40,41,42] Although PXRD data do not show evidenceo fN a 2 O, 23 Na MAS NMR experiments evidenced the formation of Na 2 CO 3 during the urea treatment, which might result from the reactiono fN a 2 Ow ith CO 2 originated from the hydrolysis of fulminic acid, as will be discussed in detail later.S odium titanate materials have been reported to exhibit Na and/or On on-stoichiometry,w hile retaining their originalc rystal structure. [43][44][45] However,b eyond ac ertain level of non-stoichiometry,t he structure of these materials become unstablea nd suffer partial decomposition. This could explain the formation of Na 2 Ti 6 O 13 in samples 10-20U. As imilarp henomenonw as observed for LiV 3 O 8 annealed at 450 8Ci nareducing atmosphere (5 %H 2 /Ar). [46] Despite the formation of a non-stoichiometric sodium titanate phase in sample 5U (see EPR data, Figure 2a)i tw as not possible to identify the Na 2 Ti 6 O 13 phase in the PXRD pattern. This could be due to the low amount of urea mixed with the as-synthesised NTO (5 wt %), whichm ight yield as mall amount of Na 2 Ti 6 O 13 that is below the detection limit of the X-ray diffractometer.W hen tested as aSIB anode,Na 2 Ti 6 O 13 showedhigherNa + ion mobility than NTO duet oi ts 3D tunnel structure but was no further considered for practical applicationsd ue to its low theoretical capacity( 49.5 mAh g À1 ). [47,48] Nevertheless, previous studies have shown that as uitable hybridisation of NTO and Na 2 Ti 6 O 13 can improve the overall anode rate performance with respectt oN TO alone. [48,49] This might suggest that the forma- tion of this phase could offset respective drawbacks stemming from its structure and enhancet he overall Na-storage behaviour.T he formation of Na 2 Ti 6 O 13 and its effecto nt he anode performance will be further discussed in the theoretical section. Field emission electron microscopy (FESEM) images of the as-prepared samples ( Figure S3) show that their microstructure consist of elongated particles with several microns in length and some hundreds of nanometresi nw idth, in agreement with literature reports on Na 2 Ti 3 O 7 . [3,11] Despite retaining the morphology after the urea treatment, ab roader particle size distribution is observed in those samples where ah igherc oncentration of urea was used. Energy-dispersive X-ray analysis (EDX) shows ah omogeneous elemental distribution of Na, Ti and Oa toms across the particles in all the samples ( Figure S4). Furthermore, combined thermogravimetric analysis (TGA, Figure S5) and transmission electron microscopy (TEM, Figure S6) suggest that the samples contain % 9wt. %o fc arbon( formed from sucrosep yrolysis), which is found as au niform coating layer with at hickness of 15-17 nm on the surface of the sodium titanateparticles.
Electron paramagnetic resonance (EPR) spectroscopy was used to unambiguously detect the presence of unpairede lectrons trapped in the structure, including the formation of oxygen vacancy sites. [50] Figure 2a shows the EPR spectrao f samples 0-20U.The EPR signal at 350 mT (g = 2.003) is attributed to the presenceo fo xygen vacancies in NTO, as reported in the literature. [9,51] Furthermore, ab road signal with am inimum at 360 mT (g = 1.95) may be explained with the presenceo f paramagnetic Ti 3 + ions. [52,53] The calculations used to obtain both g-values can be found in the Supporting Information (Eq. S1). From these data, it is observed that more oxygen vacancies and Ti 3 + ions are generated with increasing urea con-tent, as reflectedb yt he increasei nt he intensity of the correspondingr esonance peaks.
The surface chemical states of the samples were investigated by X-ray photoelectron spectroscopy( XPS), as shown in Figures 2b and S7. The high-resolution Ti 2p XPS spectra (Figure S7 b) show peak doublets at 459.5 eV and 464.5 eV,w hich correspond to the binding energies of the Ti 2p 3/2 and 2p 1/2 peaks of Ti 4 + ions, respectively. [54] No peaks relatedt ot he presence of Ti 3 + ions are observed in these data. This suggests that the concentration of Ti 3 + ions on the surface of the ureatreated samples is very diluted. These data are in agreement with previous XPS reported datao nh ydrogenated NTO samples, where an almostu ndetectable concentration of Ti 3 + ions was observed. [9,10] The high-resolution O1sX PS spectra (Figure 2b)s howapeak centred at 530.2 eV (red) which is attributed to TiÀOb onds, whereas the broader and less intense peak centred at 531.7 eV (black) is attributed to TiÀOH bonds. [10,54] TiÀOa nd TiÀOH peaks in sample 20U decrease and increasei ni ntensity,r espectively,c ompared to the pristine material (0U), suggesting that more oxygen vacanciesand hydroxyl groupsa re generated after the urea treatment. [10] Furthermore, XPS analysis does not show any evidenceo fn itrogen present in the samples as ar esult of the thermal decomposition of urea and/or the direct reaction between the samples and the N 2 atmosphereu sed during the heatingt reatment ( Figure S7 a). Therefore, it is possible to rule out the formation of aN -modified NTO compound. 23 Na MAS NMR was performedo na ll the samples (Figures S8  and S9) to further probe their local structure.T he presence of Ti 3 + speciesw ould be expectedt or educe the 23 Na spin-lattice relaxation time (T 1 )d ue to paramagnetic relaxation. [55,56] The 23 Na MAS NMR spectrum of NTO ( Figure S8 a) exhibitst wo main signals centred at 3.5 ppm and À12 ppm, which corre- spond to the Na(1) (coordination number = 7) and Na(2) (coordination number = 9) ions, respectively. [57,58] Samples 10U and 20U show peaks centred at 6.3 ppm and À18.6 ppm whichi ncrease with urea content and correspond to Na 2 CO 3 and Na 2 Ti 6 O 13 ,r espectively (Figure S8 b). [59,60] Ap ossible mechanism for the formation of Na 2 CO 3 is proposed later.A so bserved in Figure S8 a, the Na1 and Na2 line shapes do not change across the different samples. This indicates that defects might be formed either on the surfaceo ra tabulk concentration which is too low to promote any globalc hanges in the local Na coordination environments. Due to the overlap of the Na 2 CO 3 and Na 2 Ti 6 O 3 with the Na 2 Ti 3 O 7 resonances,i tw as not possible to accurately measuret he T 1 relaxation time. However,T 1 was estimatedb yi ntegrating over as mallr egion of the Na1 resonance. From samples 0-20U, the estimated T 1 relaxation time decreaseds lightly from 5.6 to 4.8 s( Figure S9). This decrease could reflect the presence of Ti 3 + defects although the small size of the reduction suggestst hat the concentrationo ft he defects must be very low.
In conclusion, based on our experimental evidence, we propose that upon heatingt he NTO and urea mixture, the latter decomposes into ammonia (NH 3 )a nd fulminica cid (HCNO)a t T % 180 8C( [Eq. (4)]), [21] and then ammonia further decomposes into reactive H 2 and inert N 2 at T > 400 8C([Eq. (5)]): [21] CH 4 N 2 O ðsÞ ! NH 3ðgÞ þ HCNO ðgÞ ðT % 180 CÞ ð4Þ The reducing H 2 gas then removes oxygen atoms from NTO creatinga nionic vacancies in the structure, as observed with EPR and XPS ( Figure 2). The process can be regarded as [Eq. (6)] [61] 2O 2À ! O 2 þ 4e À ð6Þ Upon oxygen loss, H 2 combines with O 2 to form H 2 O vapour [62] and the liberated electrons are transferred to the empty 3d levels at the bottom of the conduction band belonging to the adjacent Ti atoms. [63] Consequently,T i 4 + ions are reduced to Ti 3 + ions (as seen in the EPR data (Figure 2a)), generating unpaired electrons in the 3d shell of the Ti atom. The overall reaction process may be seen as [Eq. (7)]: [62,64] H 2ðgÞ þ Na 2 Ti 3 O 7ðsÞ ! H 2 O ðgÞ þ Na 2 Ti 4þ 3Àx Ti 3þ Furthermore, PXRD and 23 Na MAS NMRd ata ( Figure 1a nd Figure S8) show the formation of the Na 2 Ti 6 O 7 secondary phase, which is consistentw ith the creation of defects in the NTO structure. [10,40] In addition, the 23 Na MAS NMR data show the formation of Na 2 CO 3 ( Figure S8), which may be explained by two differentr eactionm echanisms; HCNO is reported to decompose to NH 3 and CO 2 when H 2 Oi sp resent in the mediumthrough ah ydrolysis process [Eq. (8)]: [64,65] HCNO ðgÞ þ H 2 O ! NH 3ðgÞ þ CO 2ðgÞ ð8Þ Since H 2 Oi sf ormed upon reduction of NTO (Eq. 7), it is likely that this reaction( Eq. 8) occurs.T he generation of CO 2 could result in the formation of Na 2 CO 3 ,e ither by the reaction of Na 2 O( formed upon decomposition of NTO) with CO 2 ,o r through an exchange of Na + in NTOw ith H + ions from H 2 O. [57] Duringt he process, Na + reacts with the OH À group of H 2 O, which in turn furtherreacts with CO 2 to form Na 2 CO 3 . [57] Electrochemical testing in Na-ion batteries Electrochemical tests including galvanostaticc harge/discharge (GCD) cycling, cycling voltammetry (CV) and electrochemical impedance spectroscopy (EIS) measurements were carried out to investigate the electrochemical properties of the as-prepared electrodes. Figure S10 showst he galvanostatic charge/ dischargev oltage profiles of samples 0-20U in the voltage range 0.01-2.5V vs. Na + /Na during cycles 1a nd 2a t0 .1 C (17.7 mA g À1 )c ycling rate. Almosti dentical load curves were observed for all the samples, except for an additional slope at 0.8-0.6 Vv s. Na + /Na (marked with an arrow in Figure S10) in sample 20U, attributed to the insertion of Na + ions into the Na 2 Ti 6 O 13 impurity,o bserved with PXRD and 23 Na NMR (Figure 1a nd Figure S8). [49,66] The absence of this slope in the load curves of sample 10U may be attributed to the lower content of Na 2 Ti 6 O 13 found in this sample.
During the first discharge process, the plateau at 0.6 Vv s. Na + /Na corresponds to the irreversible reaction of Na + ions with the carbon additive. [6,11] Furthermore, the two plateaux at the voltage below 0.2 Vv s. Na + /Na are related to two-phase transitions corresponding to Na 2 Ti 3 O 7 !Na 3Àx Ti 3 O 7 and Na 3Àx Ti 3 O 7 !Na 4 Ti 3 O 7 reactions, [7] which correspond to the insertiono ft wo Na + ions with concomitant reductiono fT i 4 + to Ti 3 + (C theoretical = 177 mAh g À1 ). [6,11] Duringt he first charge process, as inglep lateau at 0.4 Vv s. Na + /Na is observed, which corresponds to the extraction of Na + from the Na 4 Ti 3 O 7 structure. [6,11] From the second cycle onwards, only the plateau at 0.2 Vv s. Na + /Na is observed upon discharge, suggesting that the Na 2 Ti 3 O 7 !Na 3Àx Ti 3 O 7 pathway process no longer occurs and, hence,N a 2 Ti 3 O 7 transforms directly to Na 4 Ti 3 O 7 . 7 Both NTO and urea-treateds amples show av ery similar first discharge and charge capacities of ca. 420 mAh g À1 and 260 mAh g À1 ,r espectively,w ith al ow coulombic efficiency of ca. 60 % ( Figure S10). The poor coulombic efficiency observed in the first cycle is explained by the irreversible reaction of Na + ions with carbon( ca. 0.6 Vv s. Na + /Na) [6,11] and the formation of an SEI layer,w hich starts to form at voltages below 1.0 Vv s. Na + /Na. [67,68] In subsequentc ycles,t he coulombic efficiency increasest om ore than 98 %a nd remains constant up to cycle 100. The long-term cycling performance of the 0-20U electrodes at the 0.1 Cc ycling rate ( Figure S11) suggests as imilar degradation pathway over 100 cycles.F urthermore, it should be noted that the presence of Na 2 CO 3 ( Figure S8)d oes not seem to have ad etrimental effect on the overall electrochemical stability of samples 10U and 20U. These results are in accordancew itht he data reported by Ts iamtsouri et al.,w ho showed that Na 2 CO 3 did not have an effect on the cycling stability. [58] Overall, ac apacity retention upon charge of ca. 40 %w as observeda fter 100 cycles. This could be the resulto ft he formation of an unstableS EI layer and the continuous increaseo f the charge-transferr esistance, commonly observed in previous studies. [17,69,70] These data suggest that the generation of defects in the crystal structure and the presence of Na 2 Ti 6 O 13 do not lead to an improvement in the electrochemical performance at relativelyl ow cycling rates. By contrast, galvanostatic cycling at 1C (177 mA g À1 )a nd 2C(354 mA g À1 )c ycling rates (Figure 3a,b) show am ajor improvement of the capacity with increasing content of urea. Thus, the 20U sample exhibits the beste lectrochemical performance with an initial discharge capacityo f3 16 mAh g À1 (1 C) and 272 mAh g À1 (2 C) (cf. 281 (1 C) and 210 mAh g À1 (2 C) for 0U). 5U and 10U show initial capacities of 299 (1 C), 238 mAh g À1 (2 C) and 306 (1 C), 252 mAh g À1 (2 C), respectively.B yt he end of cycle1 00, 20U shows ad ischarge capacity of 154 mAh g À1 (1 C) and 145 mAh g À1 (2 C), in contrast to 106 mAh g À1 (1 C) and 90 mAh g À1 (2 C) for 0U. Figure 3c shows the rate performance of the samples at different cycling rates. At all rates, the 20U performs better than the other samples.A tt he high rates of 2, 5a nd 10 C, the discharge capacities are 160, 112a nd 80 mAh g À1 ,r espectively. Once the rate decreases from 10 Ct o0 .1 C, the discharge capacity of all the electrodes increases significantly,r eaching approximately 80 %o ft heir initial capacity.T he capacity is kept fairly constant in the following five cycles, suggesting good structurals tability and reversibility. [9,16] In conclusion, the galvanostatic data show that the urea treatment significantly increasest he rate capability and specific capacities of the active materials, leading to enhanced electrochemicalp erformance. The improvement is due to as ynergistic combination of defects generated in NTO (oxygen vacancies, Ti 3 + and Ti-OH species) and the formationo fN a 2 Ti 6 O 13 which enableh igh Na + ion/e À mobility.These results outperform many of the NTO systems reportedi nt he literature (Table S3).
To gain further insighti nto the electrochemical behaviour of the different materials studied, CV and EIS measurements were conducted. Figure S12 shows the CV curves for samples 0-20U at various scan rates from 0.05 to 0.3 mV s À1 in the voltage range 0.01-2.5 Vv s. Na + /Na. In at ypical CV for NTO at al ow scanningr ate, the first cathodic sweep involves the reactiono f Na + ions with the carbon additive at 0.36 Vv s. Na + /Na and the insertion of Na + into the layered NTO structure at 0.08 V vs. Na + /Na, concomitant with the reduction of Ti 4 + to Ti 3 + ions. [11,71] Moreover,e lectrolyte reduction starts to take place at voltages below 1.0 Vv s. Na + /Na and becomes more severe at deep dischargev oltages (0.01 V). [67,68] Upon charge, only the anodic peak related to the extraction of Na + ions from the NTO structure is observed. [11,71] As the scanningr ate increases, the anodic and cathodic peaks shift towards lower and higher voltages, respectively,s uggesting that the insertion/extraction of Na + ions becomes more sluggish. [54] Furthermore, the oxidation current peaks are higherthan the reduction peaks, indicating that the extraction process of Na + ions is more facile than the insertion process.
In ad iffusion-controlled process, the peak current is proportional to the square-root of the scanning rate. [72] Figure S13 shows al inear relationship between these two parameters, which confirms the intercalation process in all the samples to be diffusion-controlled.F urthermore, the diffusion coefficient of Na + ions, D Na þ ,c an be determined using the Randles-Sevcik relationship (Eq. S2).T he calculated intercalation D Na þ increases with increasing defectc ontent (i.e. 0U (1.2 10 À11 cm 2 s À1 ), 5U (1.7 10 À11 cm 2 s À1 ), 10U (2.7 10 À11 cm 2 s À1 )a nd 20U (4.1 10 À11 cm 2 s À1 )). Mott-Schottky plots obtained from EIS recorded at af requency of 500 Hz (Figure 3d)p rovidee vidence of higherc harge carrierd ensity due to the increased defects induced by the urea treatment.T hese are calculated to be 6.  Figure 3e (inset), which is composed of a R s at high frequencies, ac harge-transfer resistance( R CT )a long with ac onstantp hase element (CPE), a Warburg impedance (Z W )a nd ac apacitance (C). The intercept at the Z' axis in the high-frequency range (R s )i sd ominated by the resistance of the electrolyte to ion transport; the highmediumf requency semicircle (R CT )r efers to the charge-transfer resistance for electrons and Na + ions across the electrode-electrolyte interface; [73] the CPE corresponds to the electrical double layer capacitor on the interface between the electrode and electrolyte;a nd the slopeo bserved in the low-frequency domain is attributed to the Warburg impedance (Z W )a nd corresponds to the Na + ion diffusion in the bulk of the electrode, followed by the chemical capacitance of the electrode (C)a t very low frequencies. [73] The estimated charge-transferr esistances (R CT )a re shown in Ta ble S4.I nt he OCV state, the R CT of 20U is 40.0 W,m uch lower compared to 0U (i.e. 104.2 W). This suggests that the charge transfer process is enhanced when defects are generated in the sodium titanate structure, in accordance to the enhanced electrochemical performance observed at high cycling rates (Figure 3a-c). During discharge,t he electrolyte starts to decompose at voltages below 1.0 Vv s. Na + / Na, [67,68] leading to the formation of an SEI layer on the surface of the electrode, as observed in the CV measurements ( Figure S12). As the voltage further decreases, the decomposition continues to take place, leading to the formation of at hicker and more resistive SEI layer (Table S4). [69] As ar esult,a na dditional R SEI CPE SEI in series with R CT CPE CT was added to the equivalent circuit as depicted in Figure 3f.When the voltage decreases from 1.0 to 0.4V vs. Na + /Na, the semicircle progressively contracts,s howing charge-transfer resistances of 28 the SEI layer resistance is observed, confirming that the reduction of the electrolyte becomes more severe as the voltage decreases. These resultsd emonstrate an improved charge transfer process occurring in the urea-treated samples which may explain their enhanced rate capability (Figures 3a,c).

Ab initio calculations
To further understand the origin of the enhanced electrochemical rate performance of samples 5-20U, we carriedo ut two different computational analyses which correspond to the determination of the phase stability of NTO under urea treatment The stabilityo fN TO is first analysed by comparingi ts thermodynamic stabilityw itht hose of other stable phases reported for the Na-Ti-O system. This is achieved by obtaining stable atomic configurations of NTO and 19 stable compositions of Na x Ti y O z , [31] (Table S1)f ollowed by the calculation of the formation enthalpies of the obtained phases (see Experimental Section). Figure 4a shows the Na-Ti-O phased iagram at 0K,c onstructed based on the PBEsol calculated formation enthalpies. The analysisi nF igure 4a illustrates that NTO is unstable and decomposes into phases on at ie line between Na 2 Oa nd TiO 2 . The tie line includes Na 2 Ti 6 O 13 ,w hich seemingly explainst he formation of the Na 2 Ti 6 O 13 secondary phase during the urea treatment ( Figure 1a). However,aclose examination on the ternary phased iagram reveals that Na 2 Ti 6 O 13 is also thermodynamically unstable and, hence,n ot likely to be formed upon the decomposition of NTO. This discrepancyb etween the calculationsa nd experiments may arise from the differencei n temperatures used in the calculations (0 K) and in the urea treatment (723 K).
To analyse the phase stability during the urea treatment, Gibbs free energy of formation is calculated at elevated temperatures (300, 720, and 970 K). Figure 4b shows the Gibbs free energies of formation of seven phases on the tie line between Na 2 Oa nd TiO 2 .A st he temperature increases, calculations start to place the free energy of Na 2 Ti 6 O 13 below the convex hull. This suggests that Na 2 Ti 6 O 13 becomes thermodynamically stable at elevatedt emperature, allowing the decompositiono fN TO into Na 2 Ti 6 O 13 during the ureat reatment. Further calculations revealt hat, under urea treatment at 450 8C (= 723 K), the decomposition of NTO can cause changes in the Gibbs free energy (DG)f rom À6.7 to 193.4 meV atom À1 (Figure S14). The calculated DG reveals that most decomposition processes where Na 2 Ti 6 O 13 forms require energy smaller than the thermale nergy imposed during the urea treatment, that is, kT = 62.3 meV atom À1 ,w here k is the Boltzmann constant and T is the absolute temperature (723 K). This further supports the formation of Na 2 Ti 6 O 13 during the urea treatment as observed in the PXRD and 23 Na NMR data (Figure 1a nd Figure S8). However,t he above thermodynamic analysis neglects atomicd isplacements and associated kinetic barriers that also play an important role in the actual decomposition processes. To consider the kinetic contributions, we compared the atomic configuration of NTO with the other phases of the Na-Ti-O system ( Figure S15). The comparisons show that, among the various phases in the Na-Ti-O system, Na 2 Ti 6 O 13 has the most similar atomic configurations compared to NTO. [10,39,40] Thus,i ti st he most favourable phase with the lowest kinetic barriers. The phase transitionf rom NTO to Na 2 Ti 6 O 13 is particularly more likely to occur for the experiments performed under N 2 /H 2 atmosphere, because the reactive H 2 gas tends to remove Na and Of rom NTO 40 which,i nt urn, facilitates the atomic movementsf or the phase transition from NTO to Na 2 Ti 6 O 13.
Having identified Na 2 Ti 6 O 13 as the most plausible phase formed duringt he urea treatment, we next focussedo nt he effect of Na 2 Ti 6 O 13 on the electrochemical performance in samples 5-20U. One effective way to do this is to analyse the band structures of the constituents Na 2 Ti 3 O 7 and Na 2 Ti 6 O 13 that affect the electrical conductivity of anodes. Previous DFT works have calculated the band structure and associated electronic properties of NTO. [9,74] However, all calculations to date were performed based on semi-local exchange functionals such as GGA and LDAs, which largelyu nderestimate the fundamental bandgaps, [75,76] inhibiting an accurate assessment of the electronic properties of NTO. To effectively eliminate the bandgap underestimation error,w ecarriedo ut band structure calculations for NTO and Na 2 Ti 6 O 13 using an HSE06 hybrid functional that provides am ore accurate descriptiono ft he electronic structures of Ti-containing materials (Figures 5a,b). [77][78][79][80][81] Overall, the band structures of both materials are characterised by a wide bandgapw ith flat CBM, indicating that both phases are of electronic insulating nature. The flat CBM is more evident for NTO, rangingf rom the g high symmetry point (G)t oYand Bh igh symmetry points (Figure5a). Ac lose examination of the CBM reveals that the direct transition of G!G in NTO has an energy just 1.9 and 6.6 meV smaller than the indirect gap transitions of G!Ya nd G!B. Such flat CBM can render NTO to undergot he direct-indirect bandgap transition under minor changes in atomic structures, which makest he type of bandgap of NTO rathercontroversial, as usually reportedi nt he literature. [3,74,82,83] Another finding is that the fundamental bandgap of Na 2 Ti 6 O 13 (4.35 eV) is relativelyl ower than that of NTO (4.47 eV),w hich could partially explain the decrease in the bandgapo fN TO after the ureat reatment, as observed by diffuse reflectance UV-VIS spectroscopy (Figure S16), where a bandgapof3 .63 eV in the 20U sample was obtained compared to 3.88 eV for pristine NTO. In the case of ac omposite of NTO and Na 2 Ti 6 O 13 ,t he effective bandgap measured duringt he experiments can be furtherl owered, as the energetic alignment of the VBM and CBM in each phase also affects the size of the effective bandgap. [77] To test the band alignmento fN TO and Na 2 Ti 6 O 13 ,w ep rojected the band structures onto atomic orbitals of Na,T ia nd Oa nd then aligned them with respectt ot he vacuum level, as illustrated in Figures5c,d (see Experimental Section). Calculations predict the VBM and CBM of Na 2 Ti 6 O 13 to be lower than NTO by 0.07 and 0.19 eV,r espectively,c ausing an overall band structure of Na 2 Ti 6 O 13 to lie below NTO. This alignment in band structures further decreases the effective bandgapa tt he interface such that the bandgapo ft he NTO composite anode is reduced by 0.19 eV,c ompared to NTO. The decreasei nb andgapi si ng ood agreement with the UV-vis spectroscopy analysis( Figure S16), whichs howed ad ecrease of the bandgap by 0.25 eV after the ureat reatment of NTO. Further examination of Figure 5c confirms that the conduction band of both phases is mostly composed of Ti do rbital, which is consistent with previous literature. [74,84] This indicatest hat, when Ov acancies are generated in NTO structure, the resulting excess of electrons from the Ov acancy will be centred on Ti cations. This can cause the reductiono fT i 4 + to Ti 3 + and, in turn, further decrease the bandgap of NTO anode.T he decrease in bandgap could result in better electronic conductivity and, consequently,i na ne nhancement of the rate performance. In conclusion, the above calculations suggest that the enhanced anode performance of the urea treated samples is attributed to the synergetic effect of the Na 2 Ti 6 O 13 secondary phase and Ov acancy generated upon urea treatment.