Transport Properties of Nanostructured Li₂TiO₃ Anode Material Synthesized by Hydrothermal Method

Abstract

Li₂TiO₃ nanopowders were synthesized by hydrothermal process using anatase TiO₂ and LiOH·H₂O as raw materials. Li₂TiO₃ crystallizes in the layered monoclinic structure (space group C2/c) with average crystallite size of 34 nm. Morphology, elemental composition and local structure of products were carried out using high-resolution transmission electron microscopy, field-emission scanning electron microscopy, Raman and Fourier transform infrared spectroscopy. Transport properties investigated by d.c. (4-probe measurements) and a.c. (complex impedance spectroscopy) show the activation energy of 0.71 and 0.65 eV, respectively. The ionic transport properties of Li⁺ ions in nanocrystalline Li₂TiO₃ characterized by cyclic voltammetry and impedance spectroscopy validate the good electrochemical properties of this anode material for lithium-ion batteries.

1. Introduction

Among all titanium-based metal oxides, various members including TiO₂ rutile, TiO₂ anatase, TiO₂(B), LiTi₂O₄, Li₂Ti₃O₇, Li₂TiO₃, LiTiO₂ and Li₄Ti₅O₁₂ are considered as potential electrode materials for Li-ion batteries (see the phase diagram in Figure 1) [1,2,3,4,5,6,7,8,9,10,11]. In these compounds, the reversible intercalation/deintercalation process of lithium ions involves the reduction of Ti⁴⁺ to Ti³⁺ that takes place in safe voltage region (>0.5 V vs. Li⁺/Li⁰). This redox potential gives faster Li⁺ charging and sustainable cyclability to the electrode because of the congruence to the lowest unoccupied molecular orbital (LUMO) of the organic liquid–carbonate electrolyte, thus avoiding the formation of electrode/electrolyte interphase [12].

Among all titanium-based metal oxides, the rock-salt lithium metatitanate Li₂TiO₃ is an important material in various energy-related applications including lithium-ion batteries and molten carbonate fuel cells, due to its higher and flatter potential profile for lithium insertion preventing the formation of Li dendrites and the decomposition of electrolyte as well. The crystal structure was determined and refined by several workers [13,14,15]. Li₂TiO₃ exists in three polymorphic structures: the metastable cubic phase α-Li₂TiO₃ (space group Fm$\overline{3}$m), which transforms above 300 °C to the stable framework β-Li₂TiO₃ and the high-temperature cubic γ-Li₂TiO₃ phase formed at 1155 °C (space group Fm$\overline{3}$m) [16,17]. Since, we are concerned with applications of the material near room temperature, attention is focused hereunder on the stable phase β-Li₂TiO₃.

The crystal structure of β-phase Li₂TiO₃ was first determined by Lang [18] and subsequently refined by Kataoka et al. [19]. It can be described by the chemical formula Li_4/3Ti_2/3O₂ or layered notation Li(Li_1/3Ti_2/3)O₂ and crystallizes in the rock-salt monoclinic structure (space group C2/c) [18]. The unit cell has twenty–four atoms with lattice parameters a = 5.0622 Å, b= 8.7712 Å and c = 9.74787 Å, α = 90°, β = 100.01° and γ = 90° [19,20,21]. Both Li and Ti cations are located in octahedral sites with three crystallographically inequivalent Li positions, namely Li(1), Li(2) and Li(3), occupying the 8f, 4d and 4e Wyckoff sites, respectively, while two non-equivalent titanium positions Ti(1) and Ti(2) reside on 4e sites. β-Li₂TiO₃ has a layered-type framework that is a stacking of LiTi₂ layers composed of Li(3) and Ti cations sharing 4e position and of single Li layers fully occupied by Li(1) and Li(2) ions [22,23].

Various synthesis methods have been reported to grow nanocrystalline Li₂TiO₃. Conventional solid-state synthesis reaction was used to produce Li₂TiO₃ with the mixture of Li₂CO₃ and TiO₂ at 900–1100 °C for 10 h to several days [21,24,25,26]. Other preparation methods such as wet-chemical methods include sol–gel technique [27], combustion method [28,29], polymer solution synthesis [30], ball milling [31,32] and extrusion–spheronisation sintering [33] that involve at least one high temperature step (over 900 °C) to obtain a well-crystallized Li₂TiO₃ phase (named LTO hereafter). To the best of our knowledge, only few reports are available on hydrothermal synthesis of Li₂TiO₃ using microsized TiO₂ powder as raw material [17]. Uniform nano-Li₂TiO₃ powder particles were prepared via a cetyltrimethyl-ammonium bromide (CTAB)-assisted hydrothermal method [34]. Unlike many synthesis methods, the hydrothermal method allows the control of shape and size of the particles by control of the rate of precipitation of the crystallized powders directly from the solution, the uniformity of nucleation, growth and aging. Expensive surfactants or template are not necessary in this method. Thus, hydrothermal synthesis is a suitable method to produce large scale crystals with high quality at a reasonable price [35].

However, β-Li₂TiO₃ is a wide gap semiconductor with an indirect band gap at Γ–C (E_g ≈ 3.9 eV) with low electrical conductivity of 10⁻¹¹ S cm⁻¹ at room temperature [35,36,37] and it was considered as a material with electrochemical inactiveness similarly to the rock-salt Li₂MnO₃ [38]. However, synthesis of nanocrystalline LTO, under the form of nanoparticles, nanorods, nanoflowers and nanofibers, provides more active surface area and shorter diffusion paths for Li⁺ ions in the structure, which makes the material electrochemically active and improves the rate capability [39,40]. Several attempts on Li₂TiO₃-based composites as cathode or anode materials were reported considering the merits of lithium metatitanate, namely its excellent structural stability and improved electrochemical performance for Li-ion batteries [41,42,43,44,45,46,47]. The Si/LTO nanocomposite synthesized by sol-gel process as an anode material exhibited a specific capacity of 471 mAh g⁻¹ even after 50 cycles. In this nanocomposite, the structural stability is due to Li₂TiO₃, which acts as excellent buffer to the Si powders and partly compensate the capacity loss [44]. The composite 0.55Li₂TiO₃-0.45LiCrO₂ prepared by spray pyrolysis exhibits the highest initial discharge capacity of 203 mAh g⁻¹ [45]. Due to its three-dimensional (3D) path for Li⁺-ion migration and its excellent structural stability, LTO has served as both a Li ion conductive layer and a protective coating layer of cathode material against the attack of HF in the electrolyte [48,49,50,51,52].

In previous works, we have investigated the electrochemical properties of pristine nano-LTO [53], and the improve performance of LTO/CNT and LTO/graphene composites [54]. The effect of nanosized Li₂TiO₃ was evidenced by the excellent cyclability and the stable specific discharge capacity of 113, 143 and 150 mAh g⁻¹, respectively, retained over 30 charge–discharge cycles at 1C rate. In the present study, we report additional measurements that evidence the beneficial effect of nanosized scale on the transport properties of LTO specimens prepared by hydrothermal synthesis after heat treatment at 800 °C in ambient conditions. The transport properties are investigated using d.c. and a.c. electrical conductivity, cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS). Finally, the structural stability and the cyclability of LTO nanopowders cycled in aqueous cell is discussed.

2. Materials and Methods

2.1. Synthesis

Nanocrystalline β-Li₂TiO₃ was synthesized by hydrothermal reaction using 1.198 g (0.015 mol) TiO₂ powder (anatase phase, Sigma-Aldrich, USA) and 1.259 g (0.03 mol) LiOH∙H₂O (Sigma-Aldrich, USA) as raw materials. TiO₂ and LiOH∙H₂O powders (molar ratio 2:1) were added into a Teflon lined autoclave (50 mL) together with 15 mL distilled water and heated at 200 °C for 10 h. During the reaction period the autoclave was kept at the autogenous saturation vapor pressure of the solution without agitation. Then, the product was filtered, rinsed with deionized water. The as-prepared powder was dried at 50 °C in ambient conditions. The obtained powder was calcined at 800 °C for 2 h in air.

2.2. Characterization

The structure of samples was investigated by the X-ray diffraction (XRD) technique using a Siefert computerized X-ray diffractometer (model 3003-TT) equipped with a CuK_α radiation (λ = 1.5406 Å) source filtered by Ni film operated at a scan speed of 0.03 degree per second in the 2θ range 10°–70°. Raman scattering (RS) spectra were recorded at room temperature in the range of 50–1000 cm⁻¹ with a spectral resolution of ≈1 cm⁻¹ using a HR800UV Raman spectrometer (Jobin-Yvon, Longjumeau, France) using air-cooled frequency doubled Nd:YAG laser 532 nm with an excitation source of 50 mW. The spectral calibration was carried out using the 520 cm⁻¹ scattering line of silicon. A 100× objective lens was used to focus the laser light on sample surface. The Fourier transform infrared (FTIR) measurements were carried out using a VERTEX 80v (Bruker, Karlsruhe, Germany) in the spectral range 100–1000 cm⁻¹. High resolution transmission electron microscope (HRTEM) (model FEI, TECHNAI G²-30 S-twin D905) was used to analyze the microstructure of the sample. The surface morphology of the sample was observed by a field emission scanning electron microscope (FESEM, SIRION 200). The composition of the sample was analyzed by energy dispersive spectroscopy with FESEM (FESEM, SIRION 200).

A.C. conductivity was performed by impedance spectroscopy using a phase sensitive multimeter (model PSM 1700, Newton, Leicester, UK) at various temperatures from 300 to 500 °C. The amplitude of the A.C. signal was 50 mV over the frequency range 1 Hz–1 MHz. The synthesized powder was pressed into 12 mm diameter pellet with the thickness of 1.25 mm and annealed at 600 °C for 2 h to enhance the packing density. The pellet was used for electrical studies. Cyclic voltammetry (CV), galvanostatic charge–discharge experiments were conducted to evaluate the specific discharge capacity of the electrode as reported elsewhere [54]. A conventional three-electrode glass cell (Pt/saturated Li₂SO₄ aqueous solution/Li₂TiO₃) was used with an Ag/AgCl reference electrodes.

3. Results

3.1. Elemental and Structural Studies

The elemental analysis of the as-prepared Li₂TiO₃ powders sintered at 800 °C for 2 h carried out by energy dispersive spectroscopy (EDS) gives the composition of 74.5 and 25.5% (accuracy of ±0.5%) of titanium and oxygen, respectively (see Figure 2 in Ref. [54]). The LTO samples have a white color, which indicates a stoichiometric composition, according to Hoshino et al. [55] for products sintered in oxidizing conditions.

Figure 2b presents the XRD pattern of Li₂TiO₃ sample recorded using the Bragg–Brentano geometry. All the X-ray diffraction lines can be readily indexed to the monoclinic β-Li₂TiO₃ structure (JCPDS Card No. 33-0831). The XRD diagram exhibits predominant (002) reflection at 2θ = 18.48° along with other characteristic orientations ($\overline{1}$31), ($\overline{1}$33), (043), (006), (312) and (062) corresponding to monoclinic β-Li₂TiO₃ structure with C2/c space group (ordered phase). In the hydrothermal reaction, the Li-Ti-O framework is formed, characterized by the basic ($\overline{1}$33) lattice plane. The strong intensity of ($\overline{1}$33) diffraction peak gives evidence of the good ordering of lithium and titanium ions in the LiTi₂ slab of the monoclinic structure in which Li, Ti and O atoms are arranged in the sequence Li_interslab(Li_1/3Ti_2/3)_slabO₂. In this lattice, lithium ions occupy 1/3 of 4e sites, while the titanium ions occupy 2/3 other 4e sites (Wyckoff notation) [37]. Note that the intensity of the (002) supercell XRD peak is enhanced by the sintering process at 800 °C for 2 h. Rietveld refinements of XRD data were performed by the FULLPROF program [56]. The comparison between observed (dots) and calculated (solid line) patterns is shown in Figure 2 together with their difference curve. The lattice parameters of the monoclinic structure (space group C2/c) are a = 5.069(1) Å, b = 8.799(1) Å, c = 9.759(2) Å and β = 102°. The elementary unit volume (V = abc sinβ) is 429.1 ų. These results match well with the literature data and confirm that the specimens are single Li₂TiO₃ phase [15,16,17]. The crystallite size of the prepared sample was estimated using Debye–Scherrer formula [57]:

$$L=\frac{K\mathsf{\lambda }}{\mathsf{\beta }\cos \mathsf{\theta }}$$

(1)

where β is the full width half maxima (FWHM) in radians, λ is wavelength of the X–ray, θ is corresponding Bragg angle and K is a dimensionless shape factor (K = 0.94). The estimated crystallite size is 34 ± 2 nm with a lattice strain 0.009 in the rock-salt Li₂TiO₃ powders.

The short-range structural properties of the prepared samples were further investigated by Raman and FTIR spectroscopy that are considered to be powerful for either detecting impurities or determine the local environment of oxygen atoms. In lithiated oxides, where Li ions occupy octahedral sites, the frequency of Li-O stretching is known to be observed within the 200–400 cm⁻¹ spectral region, while for Li tetrahedrally coordination, the frequency lies in the 400–550 cm⁻¹ region [58]. Figure 3a shows the Raman spectra in the spectral range 100–900 cm⁻¹of several titanate oxides, i.e., TiO₂ anatase, TiO₂ rutile, Li₄Ti₅O₁₂ spinel and β-Li₂TiO₃, which evidence the unicity of the monoclinic structure of Li₂TiO₃. Among the 15 allowed Raman active modes (7A_g + 8B_g) of the C_2h⁶ spectroscopic symmetry, only 11 observed Raman bands are observed, which can be attributed as followed. The high-wavenumber peaks at 575 and 668 cm⁻¹ are assigned to the Ti-O stretching vibrations in TiO₆ octahedra; the lattice modes, i.e., A_g(T) translational lattice mode, and O-Ti-O bending modes appear at ν < 320 cm⁻¹; the low-frequency peak at 98 cm⁻¹ (with very weak intensity) is due to the Li-O stretching modes in LiO₄ and LiO₆ are observed at 358 and 430 cm⁻¹, respectively. Frequency and intensity of Raman bands are listed in

Table 1. Frequency (in cm⁻¹) and strength of Raman and infrared bands of β-Li₂TiO₃. Calculated frequencies are reproduced from Wan et al. [36].

Raman				Infrared
ωexp.	Intensity *	ωcal.		ωexp.	Intensity *	ωcal.
98	vw	83	Ag
211	w	214	Ag	139	vw	143	BU
277	m	278	Bg	266	vw	263	Bu
294	m	302	Ag	357	m	355	Au
303	m	314	Bg	390	w	393	Au
350	m	360	Ag	452	w	456	Au
402	S	401	Ag	510	S	-
420	S	432	Bg	541	m	544	Bu
484	w	470	Bg	619	S	617	Au
567	w	542	Ag	642	s	634	Au
658	S	617	Ag

* w = weak, vw = very weak, m = medium, S = strong, s = shoulder.

. Our data are in good agreement with the literature [21,36,59,60,61,62,63]. Recently, Raman and infrared bands of Li₂TiO₃ have been calculated by Wan et al. [36] from first-principles total energy calculations with a generalized gradient approximation and plane-wave pseudopotential model. Our experimental data are compared with calculated frequencies in Table 2. Note that similar situation has been shown in the Raman spectrum of monoclinic Li₂MnO₃ [64,65]. All mode frequencies match well with the relation:

$$\frac{{\mathsf{\omega }}_{L{i}_{2}Ti{O}_3}}{{\mathsf{\omega }}_{L{i}_{2}Mn{O}_3}}=\sqrt{\frac{m_{Mn}}{m_{Ti}}}$$

(2)

Considering the molecular model, the Raman spectrum of the β-Li₂TiO₃ phase essentially displays stretching and bending of the various polyhedral units constituting the monoclinic lattice without appearance of Raman bands due to anatase TiO₂ (at 144 cm⁻¹) and rutile TiO₂ (at 449/610 cm⁻¹).

The FTIR spectrum of Li₂TiO₃ is shown in Figure 3b. It displays only 9 infrared bands instead of 18 expected by the group factor analysis. The two broad peaks at 510 and 619 cm⁻¹ are assigned to the antisymmetric stretching vibrations of Ti-O bonds. The low-frequency bands in the range 350–450 cm⁻¹ are a mixture of the O-Li-O and O-Ti-O bending modes due to the presence of Li ions in the LiTi₂ slabs. The band at 263 cm⁻¹ is attributed to the stretching mode of Li-O bonds in LiO₆ octahedra. The shoulder of the higher intensity peak at 642 cm⁻¹ should be related to the local distortion of the lattice that results in a decrease of the local structural symmetry generating additional infrared bands [66].

3.2. Surface Morphology

The surface morphology and microstructural features of hydrothermal synthesized Li₂TiO₃ material calcined at 800 °C for 2 h are observed by field emission scanning electron microscopy (FESEM) and high-resolution transmission electron microscope (HRTEM) in Figure 4 and Figure 4b, respectively. FESEM image of Li₂TiO₃ shown in Figure 4a reveals that the sample consists of homogeneously distributed spherical-like shaped grains with an average particle size of 150 nm without any obvious aggregations. Figure 4b–d show the HRTEM image, selected area diffraction pattern (SAED) and lattice fringes images of Li₂TiO₃, respectively. The corresponding selected-area electron diffraction (SAED) pattern of the microstructure (Figure 4c) shows that the nanoparticles are well crystallized. The d-spacing values obtained from SAED patterns are indexed to the respective planes of (002), ($\overline{1}$31), ($\overline{1}$33) and (312) that agree well with XRD patterns of the monoclinic structure of Li₂TiO₃. Figure 2d shows the layered structure with fringe spacing of 0.48 nm matching well with the d-spacing of the high intensity plane (002) of Li₂TiO₃ and corresponding to LiTi₂ slab along the c-axis [67].

3.3. Electrical Transport

Conductivity measurements were carried out using both d.c. and a.c. techniques to analyze the electrical transport of electrons and ions in the LTO lattice. Impedance spectroscopy is an efficient tool to study a.c. conductivity of materials, when using a large frequency domain, this technique provides analysis of intragrain, intergrain and relaxation process in material (Figure 5a–e). Figure 5a,b show the Nyquist plots (Z′(ω)-Z″(ω) plane) of nanocrystalline Li₂TiO₃ in a wide frequency range (1 Hz-1 MHz) over the temperature range 320–500 °C. The bulk resistance (R_b) of the LTO sample is determined from the intercept of each semicircle with the real axis. It is observed that both the intercept with the Z′ axis and the maximum of Z” shift as temperature increases. At low temperatures (T < 400 °C), the Li₂TiO₃ pellet exhibits a high resistance that is mainly due to the bulk property of the sample. At higher temperature, all the Nyquist plots consist of a slightly depressed semicircle in the high-frequency region followed by another depressed semicircle at low frequencies (<1000 Hz). The depressed semicircle, which meets real axis at high frequencies is the response of the bulk resistance (R_b) of the material, i.e., small deviation from a Debye-type relaxation process, while the depressed semicircle originated in the low frequency region is due to grain boundary (R_gb) effect. The high- and low-frequency complex impedance plots for all temperatures can be modeled using a parallel combination of resistance and constant phase element (CPE) equivalent circuit describing the bulk resistance and grain boundary components in Li₂TiO₃ sample. The bulk electrical conductivity (σ) was calculated using the formula:

$${\mathsf{\sigma }}_0=\frac{1}{R_{\mathrm{b}}}\times \frac{l}{S}$$

(3)

where R_b is the bulk resistance, l is the thickness of the sample and S is the active area of the sample.

The real Z′(ω) and imaginary Z″(ω) part of the Li₂TiO₃ sample impedance are presented for all temperatures in Figure 5c,d, respectively. The plots of Figure 5c show a decrease of Z′(ω) vs. frequency, so that σ(ω) increases with frequency (see Figure 5e). At low frequency, σ increases importantly with temperature. At high frequencies, however, Z′(ω) becomes almost temperature independent so that the Z′(ω) curves at different temperatures merge approximately in a single curve. This is due to the release of space charges caused by reduction in barrier properties of the material [38]. This unique curve at high frequency shows a dip, which is associated to charge carrier hopping in the material. In another hand, Z″ = Im(Z(ω)) reaches a maximum, which shifts towards higher frequency with temperature. This is attributed to the active conduction through the grain boundaries of the sample. The magnitude of Z″(ω) decreases with increasing temperature indicating a decrease of the resistivity of the material. The peak broadening observed with increasing temperature is attributed to a temperature dependent relaxation process in the material. The asymmetric broadening of the peaks indicates the spread of relaxation time in the sample. Our data match well with the results reported in the literature [37,38,68]. The ac response obeys the power law [69]:

σ(ω) = σ_ac = σ₀ + Aωⁿ

(4)

where σ₀ is the d.c. conductivity (at ω = 0), A is a thermally activated quantity and n the fractional constant that is 0.5 < n < 0.8 for an ionic conductor [70,71]. The frequency exponent n (Equation (4)) can be analyzed by a mechanism based on charge carrier hopping between defect sites proposed by Elliott [72]:

$$n=\frac{\partial (\ln {\mathsf{\sigma }}_{\mathrm{ac}})}{\partial (\ln \mathsf{\omega })}=1-\frac{6k_{\mathrm{B}}T}{W_{\mathrm{m}}}$$

(5)

where T is the absolute temperature, k_B the Boltzmann constant and W_m the maximum barrier height (energy of the transport charge). Using Equation (4), from the slope of curves in Figure 5d, one can derive at the highest frequency with n ≈ 0.72 and the value of W_m is 0.63 eV at room temperature.

Figure 6a,b display the Arrhenius plots of σ_ac and σ_dc, respectively. The solid lines are the fit of the temperature dependence of conductivity thermally activated according the Arrhenius-type behavior:

$${\mathsf{\sigma }}_{\mathrm{ac},\mathrm{dc}}={\mathsf{\sigma }}_{0}T\exp \left(\frac{-E_{\mathrm{a}}}{k_{\mathrm{B}}T}\right)$$

(6)

where σ₀ is the pre-exponential factor, T the absolute temperature, k_B the Boltzmann constant and E_ac the activation energy of the conductivity. The fit of the curves σT vs. 1/T in Figure 6a,b provides an activation energy E_a = 0.71 eV for σ_dc and 0.65 eV for σ_ac measured at 50 kHz. These values match well with the value 0.77 eV found for nanoparticles [37], against 0.88 eV [38] or even 0.95 eV for the bulk material [37]. Thus, E_a appears to be dependent on the crystallite size. This is corroborated by the fact that the value E_a = 0.71 eV obtained in the present work (34 nm crystallite size) is slightly smaller than that for crystallites of 88 nm reported by Dash et al. [38]. Therefore, nanosized particle favors the electronic transport and thus the ionic transport since both are correlated to maintain the electrical charge neutrality inside the material.

As shown in Figure 5c, Z″ = Im(Z(ω)) exhibits a peak at the particular frequency, i.e., the relaxation frequency, which corresponds to a single relaxation time that fulfils the relation 2πf_mτ_m = 1, where f_m is the frequency of the maximum of Im(Z). The variation of τ with T obeys an Arrhenius law given by [73]:

$$\mathsf{\tau }={\mathsf{\tau }}_0\exp \left(\frac{-W_{\mathrm{a}}}{k_{\mathrm{B}}T}\right)$$

(7)

where τ₀ is the pre-exponential factor and W_a the activation energy. Figure 6c shows the temperature dependence of the relaxation time of LTO sample in the range 320–500 °C. The mean relaxation time of the process is measured in fractions of milliseconds that does not implies electronic and ionic lattice polarization but slow relaxation can be imposed by permanent molecular dipoles, ion defects of a dipolar type, or mobile hopping charge carriers [74]. The activation energy estimated from the linear fit is found to be ~0.66 eV, which indicates a quasi-Debye behavior of the relaxation process of charge carriers in the LTO lattice. The mechanism of ionic motion has been discussed by several workers [23,74,75]. Li₂TiO₃ has a three-dimensional path for Li⁺ ion diffusion, in which the ionic migration can occur in the (00l) plane, that is, LiTi₂ layer, and along the c axis [74]. The conductivity occurs with the hopping of Li⁺ ions from tetrahedral sites to adjacent octahedral sites. At low temperatures, defects are created by surface modification. At high temperature a drift of large number of Li⁺ ions released from tetrahedral sites (LiTi₂ layer) then occupied octahedral sites (Li layer), creating interstitial site vacancies that are thermally activated and some structural disorder possibly occurred that facilitate ion transport induced by low activation energy.

3.4. Li⁺-ion Diffusivity

Investigations of the diffusivity of Li⁺ ions in the LTO lattice were carried out by cyclic voltammetry (CV) at sweep rate in the range 1–50 mV s⁻¹ and by electrochemical impedance spectroscopy (EIS) measurements in the frequency range 1–10⁶ Hz for Pt/saturated Li₂SO₄ aqueous solution/Li₂TiO₃ cells. Cyclic voltammograms display one set of well-defined current peaks (see Figure 7 in Ref. [53]) corresponding to the redox reaction in Li//Li₂TiO₃ cell, according the (de)insertion reaction:

Li₂Ti⁴⁺O₃ + xLi⁺ + xe⁻ ⇆ Li_2+xTi^(4−x)+O₃

(8)

As the starting electrode material is Li₂⁺Ti⁴⁺O₃, the insertion of the faction x of Li⁺ ions in Li_2+xTiO₃ implies the reduction of tetravalent to trivalent titanium [45]. At sweep rate of 10 mV s⁻¹, the reduction peak occurred at 0.338 V during cathodic sweep corresponding to insertion of lithium into interstitial sites of Li_2+xTiO₃ (x equals to 0.5), while the oxidation peak occurred at 0.643 V during anodic sweep. Thus, a peak separation of 0.305 V for Li₂TiO₃ is ascribed to the Li⁺ ion storage at the solid–electrolyte interface during charge–discharge process [76].

Figure 7a displays the dependency of the peak current Ip for the redox peaks as a function of scan rate ν in CV measurements (typical CV response at ν = 10 mV s⁻¹ is shown in the inset of Figure 7a). For cathodic and anodic reactions, the plots in double logarithmic scale show mutual linear behavior over the entire measurement range with a slope of 0.511 and 0.497, respectively. These values are very close to 0.5, which characterizes the conventional diffusion-controlled faradaic reaction, i.e., the lithium insertion/extraction reaction in LTO. The classical Randles–Sevcik equation for a semi-infinite diffusion of Li-ion into Li₂TiO₂ layer can be applied [77]:

$$I_{\mathrm{p}}=0.446{\left[\frac{n^3{F}^2{C}_{\mathrm{Li}}^2{A}^2{D}_{\mathrm{Li}}\nu }{RT}\right]}^{\frac{1}{2}}$$

(9)

where I_p is the peak current, n the charge-transfer number, A the surface area in cm² of the electrode, D_Li the chemical diffusion coefficient, ν the scan rate, C_Li the bulk concentration of Li-ion in electrode (~1.8 × 10⁻³ mol cm⁻³ calculated from the volume of Li₂TiO₃ (429.1 ų)), F is Faraday constant, R the gas constant and T the absolute temperature. Based on Equation (9), the apparent diffusion coefficient of Li-ion in the nanostructured Li₂TiO₂ sample was calculated to be 9.2 × 10⁻¹² cm² s⁻¹.

Electrochemical impedance spectroscopy (EIS) was carried out over the frequency range from 1 Hz to 1 MHz for Li_2+xTiO₃ electrode in the fully charged state, i.e., x (Li) = 0. The Nyquist plot consists of a depressed semicircle in high frequency and straight line at low frequency region. The interception of Z′ axis at high frequency indicating electrode/electrolyte contact resistance corresponds to the solution resistance (R_s). The depressed semicircle in the middle frequency region denotes the charge transfer resistance R_ct related to electrochemical reaction at electrode/electrolyte interface. At low frequency, the observed straight line that is the response of the Warburg impedance Z_w, which is related to solid-state diffusion of Li⁺ ions. The impedance spectrum was fitted using the Randles equivalent circuit shown in the inset of Figure 7b. Resistance R_S of the electrode is 38 Ω; R_ct is ≈250 Ω and slope of the Warburg impedance (θ ≈ 45°) indicates that the electrode is controlled by a diffusion process. In the Warburg regime, the impedance varies with the angular frequency ω according to the law [78]:

Z′ = R_s + R_ct + σ_w ω^−1/2

(10)

in which the Warburg impedance σ_w is obtained from the slope of Z′ vs. ω^−1/2 in the low-frequency range (Figure 7b). The apparent diffusion coefficient Li⁺ ions can be quantified from the low-frequency Warburg impedance according the equation [78]:

$$D_{\mathrm{L}{\mathrm{i}}^{+}}=\frac{1}{2}{\left[\frac{RT}{F^{2}A{C}_{\mathrm{Li}}{\mathsf{\sigma }}_{\mathrm{W}}}\right]}^2$$

(11)

where R, T and F are the usual constants, A the surface area of the electrode-electrolyte interface and C_Li the lithium-ion concentration in the electrode. Linear fit of data in Figure 7b gives σ_w = 9.36 V s^−1/2 and D_Li+ = 2.1 × 10⁻¹¹ cm² s⁻¹. Note that the D_Li+(EIS) value is larger by a factor of 2 in comparison to that obtained by CV. Similar difference in D_Li+ for the same cathode material by two different techniques is commonly observed [79,80].

Figure 8a,b present the cyclability of the Li_2.55TiO₃ electrode (discharge state). The XRD patterns of the starting Li₂TiO₃ and final Li_2.55TiO₃ phase after cycling process are displayed in Figure 8a displays. These results show that the structure has retained after 30 insertion-extraction cycles with the C2/c monoclinic symmetry. This insertion process is in good agreement with the model of Bian and Dong [59] who demonstrated the formation of a non-stoichiometric Li_2+xTiO₃ with x = 0.2 that retains the initial rock-salt structure. These authors have proposed three possible occupation mechanisms in Li-rich Li₂TiO₃: (i) excess of Li⁺ ions occupied the tetrahedral interstitial sites, which took the composition Li_2+4yTi_1−yO₃; (ii) the Li⁺ ions are located on the Ti⁴⁺ sites formed as antisite defects (2Li″_Ti = 3V_O^••); and (iii) excess of Li⁺ ions occupied itself site and Ti⁴⁺ and O²⁻ site vacancies formed simultaneously. The plot of the specific discharge capacity vs. cycle number for the Li₂TiO₃ electrode over 30 cycles in the voltage range 0.0–1.0 V is shown in Figure 5b. The Li₂TiO₃ anode exhibited an initial discharge capacity of ≈122 mAh g⁻¹ that is retained up to ≈114 mAh g⁻¹ after 30 cycles. The capacity retention is about 94%. In this figure the present results are compared with those of literature [41,43,46,47].

4. Discussion

In view of these previous results reported in Figure 8b, one understands why Li₂TiO₃ was not considered as a possible anode element. Indeed, the material in these works were not investigated as an anode for itself, but was used as an element entering in the composition of a composite with another anode material, usually Li₄Ti₅O₁₂ [46,47], or LiFeO₂ [41,43]. Coating lamellar cathode element with Li₂TiO₃ has also been made [81]. In any case, the voltage range used for electrochemical investigation was 1–3 V, since Li₄Ti₅O₁₂ or LiFeO₂ or the lamellar compounds are active only at voltage higher than 1 volt. However, in the range 1-3 V, Li₂TiO₃ is simply electrochemically inactive. The purpose of the addition of Li₂TiO₃ in the composite was to increase the ionic conductivity and stabilize the structure of the electrode materials [46,48,81] by taking advantage of the strong Ti-O bond in Li₂TiO₃ [82,83,84]. In addition, Li₂TiO₃ is known to be a fast Li⁺ conductor when it is doped [85,86]. Here, no doping has been made, but the conductivity has been improved owing to the nano-size of the particles. In the present work, we have shown that nanostructured Li₂TiO₃ has interesting electrochemical properties at low voltage (<1 V). This result can also explain former results obtained on Li₄Ti₅O₁₂/Li₂TiO₃ composite. In Ref. [47], the authors noticed an increased capacity of this composite, associated to a sloping discharge in the range 1.25–0 V with respect to the results observed when cycling with a lower cut-off of 1.25 V. Although this effect was attributed to an intrinsic property of Li₄Ti₅O₁₂ since an increased capacity in this voltage range has been reported by one group with Li₄Ti₅O₁₂ alone [87], the present work suggests that at least part of this effect can be attributed to Li₂TiO₃. In addition, the enhanced conductivity and the correlated improved Li-diffusion by reduction of the Li₂TiO₃ material to the nanosize observed in the present work explains the enhanced electrochemical performance of recent electrospun Li₄Ti₅O₁₂/Li₂TiO₃ composite nanofibers observed recently [88].

Table 2. Values of the activation energy for the charge carrier transport in Li₂TiO₃ reported in the literature.

Method Used	Activation Energy (eV)	Ref.
7Li NMR, a.c. conductivity	0.47–0.80	[89]
a.c. conductivity	0.81–0.91	[90]
electrical conductivity	0.60–0.90	[37]
6,7Li NMR	0.27	[74]
periodic quantum−chemical DFT	0.44–0.54	[75]
complex impedance	0.77/0.88	[38]
d.c. conductivity	0.86	[68]
6,7Li NMR	0.52	[91]
atomistic simulation	0.51	[92]
DFT calculations	0.76	[93]
a.c. conductivity	0.65/0.71	this work

Table 2. Values of the activation energy for the charge carrier transport in Li₂TiO₃ reported in the literature.

Method Used	Activation Energy (eV)	Ref.
7Li NMR, a.c. conductivity	0.47–0.80	[89]
a.c. conductivity	0.81–0.91	[90]
electrical conductivity	0.60–0.90	[37]
6,7Li NMR	0.27	[74]
periodic quantum−chemical DFT	0.44–0.54	[75]
complex impedance	0.77/0.88	[38]
d.c. conductivity	0.86	[68]
6,7Li NMR	0.52	[91]
atomistic simulation	0.51	[92]
DFT calculations	0.76	[93]
a.c. conductivity	0.65/0.71	this work

Few experimental works reported the kinetics of Li⁺ ion in the LTO framework, while the Li ion dynamics is discussed controversially. Table 2 list the values of the activation energy for the charge carrier transport in Li₂TiO₃ reported in the literature. Vijayakumar et al. utilized a combination of ^6,7Li nuclear magnetic resonance (NMR) and molecular dynamics (MD) simulations to predict that Li ions can be occupy a tetrahedral site if a Ti cationic defect is present in the LiTi₂ layer. They obtained a small activation energy of 0.27 eV for Li diffusion in β-Li₂TiO₃, which seems to be associated with impurities and stoichiometric deviation [74]. Ruprecht et al. [89] reported a self-diffusion coefficient of 2 × 10⁻¹³ cm² s⁻¹ measured by 7Li NMR stimulated echoes at 160 °C and an extremely slow Li diffusion, i.e., D_Li⁺ = 3 × 10⁻¹⁷ cm² s⁻¹ at room temperature, evaluated from a.c. conductivity with the help of the Nernst-Einsten equation. Islam and Bredow [75] determined activation energies ranging between 0.44 and 0.55 eV for Li+ ion migrating along the ab plane and perpendicularly to the LiTi₂ layers as well. The early work by Vitins et al. [90] shows that the electrical conductivity of Li₂TiO₃ is strongly dependent on the heat treatment conditions during sample preparation. Typical specimen mixed with 1.5 mol% Li⁺ and annealed at 760 °C for 20 h exhibits a conductivity of σ₃₀₀ = 4 × 10⁻⁶ S cm⁻¹ and E_a = 0.81 eV in the temperature range 187–424 °C. The authors suggested an excess TiO₂ incorporated in the lattice by Ti_Li³⁺ defects charge compensated for by V_Li⁻¹ defects, which facilitates Li diffusion. Ruprecht et al. [89] reported a dc conductivity of 5 × 10⁻¹³ and 4 × 10⁻⁶ S cm⁻¹ at room temperature and 250 °C, respectively (with E_a = 0.8 eV for long-range Li transport above 100 °C), which is very close with the value 3 × 10⁻⁶ S cm⁻¹ at 300 °C obtained by Fehr and Schmidbauer [37]. Employing atomistic simulations to study the point defect processes and lithium diffusion in LTO, Kuganathan et al. evaluated the activation energy of 0.51 eV for Li⁺ migration along the ab plane and showed that the energetically favorable intrinsic defect is lithium Frenkel type with 1.25 eV per defect [92]. More recently, Ki et al. considered three types of lithium vacancies to investigate the diffusion behavior of tritium in LTO using density functional theory (DFT) calculations and found a minimum energy barrier of 0.76 eV for the tritium atom to escape the Li vacancy well [93].

5. Conclusions

The nanocrystalline Li₂TiO₃ phase has been successfully synthesized using a simple, low cost and scalable hydrothermal reaction. The structural analyses carried out using X-ray diffraction, Raman and FTIR spectroscopy show that β-Li₂TiO₃ crystallizes with C2/c space group with crystallites 34 nm. FESEM image shows that the nanocrystalline β-Li₂TiO₃ particles are homogeneously distributed as nearly spherical shaped grains with an average size of 150 nm. The Raman and FTIR spectroscopy confirm the impurity free monoclinic structure exhibiting well-defined vibrational modes Ti-O and Li-O stretching and O-Ti-O and O-Li-O bending modes. The electrical properties revealed a thermally activated conduction mechanism (Arrhenius-type behavior) with an activation energy of 0.71 eV. The conduction mechanism of the sample is greatly improved due to more charge carriers corresponding to nanoparticles. The electrochemical of β-Li₂TiO₃ studies demonstrate that reduction of Ti⁴⁺/Ti³⁺ (0.55 Li uptake) occurred in potential range 1.0–0.0 V. The nanocrystalline β-Li₂TiO₃ shows a specific discharge capacity of 114 mAh g⁻¹ after 30 cycles at current density of 0.05 mAh cm^-2. These results suggest that nanocrystalline β-Li₂TiO₃ nano particles have shortened the Li–ion diffusion lengths and enhanced the electrochemical kinetics of the sample, which are very important keys to high rate capacity. The electrode retains 94% of capacity even after 30 cycles with an excellent structural stability.