Electrolytic Reduction of Titanium Dioxide in Molten LiCl–Li₂O

Abstract

The electrolytic reduction of TiO₂ in LiCl–Li₂O (1 wt.%) at 650 °C was investigated under a series of cathodic reduction potentials and applied charges to provide a mechanistic understanding of the electrochemical characteristics of the system. The optimal cathodic reduction potential was determined as being −0.3 V vs. Li/Li⁺. Li₂TiO₃ and LiTiO₂ were structurally identified as intermediate and partial reduction products of the TiO₂ electrolytic reduction. The reduction of LiTiO₂ was extremely slow and reversible due to its high stability and the detrimental effect of Li₂O accumulation within the solid particles. The most reduced product obtained in this study was LiTiO₂, which was achieved when using 150% of the theoretical charge under the optimal reduction potential. The highest reduction extent obtained in this study was 25%. Based on theoretical DFT modeling, a detailed multistep reduction mechanism and scheme were proposed for TiO₂ electrolytic reduction in LiCl–Li₂O (1 wt.%) at 650 °C.

1. Introduction

Titanates and titanium metal are valuable for batteries, coatings, and structural material applications [1,2,3,4]. Electrolytic reduction is a promising way for Ti metal and intermediates production directly from solid titanium dioxide through molten salt electrolysis. It has attracted extensive research effort that involves using various salt systems under various conditions with different setups [5,6,7,8,9]. The Fray–Farthing–Chen (FFC) Cambridge and calciothermic Ono–Suzuki (OS) processes are the two most well-studied processes utilizing CaCl₂ (melting point 772 °C) as electrolytes for electrolytic reduction between 850 and 950 °C. As an alternative electrolyte, LiCl can significantly decrease the reduction temperature to 650 °C for the electrolytic reduction of metal oxides due to its lower melting point [10,11,12]. The FFC process presumes that TiO₂ is reduced directly in CaCl₂ via electrons through multiple titanate intermediates, including CaTiO₃ and CaTi₂O₄, as well as titanium suboxides (TiO_x, x < 2) [13,14,15,16]. Through the OS process, TiO₂ is reduced using in-situ-generated calcium metal from the CaCl₂ electrolyte at an overpotential when the applied potential is higher than the electrolyte decomposition potential [7,8,9,17,18]. Similarly, the reduction of TiO₂ in LiCl leads to the formation of lithium titanates, including LiTiO₂ and Li₂TiO₄ when the reduction potential is lower than the LiCl decomposition potential, that is, an underpotential. The complete reduction of TiO₂ to Ti is obtained at high cell potentials [10,19,20]. Regarding the electrolytic reduction of TiO₂ in molten LiCl–Li₂O, the oxides Li₂TiO₄, LiTiO₂, and suboxides have been identified using XRD and TEM under various temperatures and potentials (

Table 1. Electrolytic reduction products for TiO₂ in LiCl salts.

Reference	Temperature	Reduction Conditions	Products	Characterization
[19]	700 °C	TiO2 pellets, cell potentials 1.8 V and 3.2 V	Li2TiO4, LiTiO2	ICP, XRD, and TEM
[20]	650 °C	TiO2 powder, constant current at 1.2 A, cell potential 2.7 V	Li2TiO4, LiTiO2, Ti	XRD and SEM
[10]	650 °C	TiO2 powder, cell potential 3.0 V	LiTiO2, TiO, Ti2O, Ti	XRD and SEM
[21]	850 °C	Thin TiO2 film, 0 to −2.0 V vs. Ni/NiO	Li2TiO3, TiO, Ti3O5, Ti2O3, Ti	CV

) [10,19,20,21]. Li₂TiO₃ was reported based on cyclic voltammetry (CV) results.

However, the TiO₂ reduction mechanism in LiCl is not as well understood as in CaCl₂. The reduction pathways and mechanism of TiO₂ in LiCl have mostly been studied using two-electrode electrochemical cells. It is unclear whether the underpotential (lower than the electrolyte decomposition potential) or overpotential electrolytic reduction was employed under these conditions [10,22,23]. To elucidate the TiO₂ electrolytic reduction mechanism in LiCl, this work systematically investigated the electrolytic reduction of TiO₂ in LiCl–Li₂O (1 wt.%) at 650 °C using a three-electrode electrolysis cell by directly controlling the cathodic reduction potential under a series of applied charges. Reduction intermediates and products were analyzed and quantified using X-ray diffraction (XRD) with Rietveld refinement analysis. The CV and chronoamperometry techniques, combined with density functional theory (DFT) calculations, were employed to obtain a detailed understanding of the TiO₂ reduction characteristics and mechanisms in the LiCl–Li₂O system.

2. Materials and Methods

A molten salt mixture, namely, LiCl–Li₂O (1 wt.%), for TiO₂ electrolytic reduction was made of ultra-high-grade lithium chloride (LiCl, 99.995% metals basis, ultra-dry, Alfa Aesar, Tewksbury, MA, USA) and lithium oxide (Li₂O, 99.5%, Alfa Aesar), which was used as received. Titanium dioxide powders (Rutile TiO₂, 99.995% metals basis, Alfa Aesar) were pelletized, grounded, and sieved between mesh 60 and 20 (particle sizes 0.25~0.85 mm). A total of 2.0 g of TiO₂ pellets were loaded in a basket for each TiO₂ reduction run immersed in a 150 g electrolyte salt bath at 650 °C.

A three-electrode electrochemical cell located in an argon glovebox (MBraun, Labmaster 200G, Stratham, NH, USA) with controlled moisture and oxygen levels (<1 ppm) was used for the electrochemical measurement and TiO₂ electrolytic reductions. The three-electrode electrochemical cell was composed of a working electrode, a Ni/NiO reference electrode, and a glassy carbon crucible, which was also used as the counter electrode (HTW GAZ30). VersaSTAT 4 Potentiostat (Princeton Applied Research, Oak Ridge, TN, USA), which interfaced with the glovebox, was employed for the electrochemistry measurement and electrolysis up to 2 amps using VersaStudio for controlling the parameters and for data acquisition. A Kerrlab electric furnace (Auto Electro Melt Maxi, Orange, CA, USA) was used to keep the electrolyte salt bath at 650 °C. This three-electrode electrochemical cell setup has shown high efficiency and reliability in a previous study for NiO electrolytic reduction [24].

Cyclic voltammetry (CV) was measured in the LiCl–Li₂O (1 wt.%) salts at 650 °C to determine the Li deposition potential ($E_{Li/L{i}^{+}}$) at a scan rate of 20 mV s⁻¹ using a Ni/NiO reference electrode and a 0.64 mm diameter stainless steel wire as the working electrode. A cathode basket assembly described in a previous work was used for the TiO₂ electrolytic reductions [24]. Briefly, the cathode basket assembly included a 1.59 mm diameter stainless steel rod as an electron collector and a stainless-steel cathode basket (height: 2.5 cm, diameter: 1 cm) made of double layers of 100 mesh stainless steel wire cloth, and the TiO₂ pellets were loaded and packed around the electron collector. The whole electrode assembly was soaked in the molten salt for 1~1.5 h before the electrolytic reduction started. Various cathodic reduction potentials, including −0.2 V, −0.3 V, and −0.4 V vs. Li/Li⁺ were applied on TiO₂, respectively, using 80% of the theoretical charge (i.e., the theoretical charge for the complete reduction of TiO₂ to Ti). The constant reduction potential was applied at an interval mode (the cell spent 8 min on and 2 min off) to minimize the excessive lithium metal generated on the working electrode. Higher theoretical charges, namely 130% and 150%, were also investigated for the TiO₂ reduction at the optimal potential, which was determined as being −0.3 V vs. Li/Li⁺ in this study for further investigation. After the TiO₂ reductions, the electrode bundles were lifted out of the molten salt and cooled down to room temperature for further treatment and product analysis. To determine the starting composition for the electrolytic reduction, a soaking test was performed by immersing the TiO₂ in the LiCl–Li₂O (1 wt.%) salt at 650 °C for 5 h without electrolysis.

The cooled cathode basket assemblies were transferred out of the glovebox and rinsed with nanopure water to remove the extra LiCl–Li₂O salts. Then, the samples were retrieved out of the basket and rinsed in acetone for a quick dry. After that, the samples were kept vacuum dried at room temperature overnight. The completely dried samples were ground into fine powders using a pestle and mortar for the X-ray diffraction analysis (Rigaku SmartLab XRD, Wilmington, MA, USA, Cu Kα radiation at 40 kV and 30 mA). The scan range (2θ) was from 10° to 80° at 4 deg min⁻¹. The XRD data were analyzed with Rietveld refinement using WPPF (whole powder pattern fitting) within PDXL 2 (Rigaku) for quantitative compositional analysis.

In this study, reduction intermediates, including Li₂TiO₃ and LiTiO₂, were identified and quantified, but LiTiO₂ was the only partial reduction product for TiO₂. Theoretically, the charge transfer for the reduction of TiO₂ to LiTiO₂ is 1, while the charge transfer for the reduction of TiO₂ to Ti is 4. The reduction extent of TiO₂ is defined as given in Equation (1):

$$\mathrm{Reduction}\mathrm{extent}=\frac{\frac{1}{4}\times m_{{\mathrm{LiTiO}}_2}}{m_{{\mathrm{TiO}}_2}+m_{{\mathrm{Li}}_2{\mathrm{TiO}}_3}+m_{{\mathrm{LiTiO}}_2}}\times 100\%$$

(1)

The total applied charge is represented as a percentage of the theoretical charge, which for the complete reduction of TiO₂ to Ti metal is given as Equation (2):

$$\mathrm{Theoretical}\mathrm{charge}=4m_{{\mathrm{TiO}}_2}F$$

(2)

where m is the number of moles of TiO₂, and F is the Faraday constant (96,485 C mol⁻¹).

The Cottrell equation was employed to estimate the oxygen ion diffusion coefficients (D) during the TiO₂ electrolytic reduction using the chronoamperometry technique. The oxygen ion diffusion coefficients were calculated based on the I(t) data fit using Equation (3):

$$I\left(t\right)=\frac{nFA{D}^{1/2}{C}_0}{{\pi }^{1/2}}\frac{1}{t^{1/2}}$$

(3)

where I is the decay current during the cell-on interval, n is the number of electrons transferred (n = 2), A is the surface area of the working electrode, C₀ is the initial concentration of oxygen ions, and t is the time in seconds.

3. Density Function Theory Modeling

Density function theory (DFT) modeling provides theoretical understanding that can be used to determine the TiO₂ reduction pathway and mechanism. Vienna Ab Initio Simulation Package (VASP) (Vienna, Austria) was used for the DFT calculation [25]. The Gibbs free energy of formation for titanate intermediates, such as Li₂TiO₃ and LiTiO₂, were calculated. The calculations of the total energies and structural relaxations were carried out using the projector augmented wave (PAW) method to model the core electrons (see the Supplementary Materials) [26]. The Helmholtz free energy, F(V,T), which is a function of both crystal volume (V) and temperature (T), can be expressed using Equation (4):

$$F\left(T,V\right)=E\left(V\right)+F_{vib}\left(V,T\right)+F_{el}\left(V,T\right)$$

(4)

where E(V) is the total energy. F_vib (V,T) and F_el (V,T) represent the vibrational and thermal electronic contributions to the free energy, respectively. The E(V) of a given crystal can be described using the equation of state (EOS), and its values at different volumes were directly obtained from periodic DFT calculations. Here, the fourth-order Birch–Murnaghan (BM4) EOS, as expressed by Equation (5), was used:

$$E\left(V\right)=a+b{V}^{-2/3}+c{V}^{-4/3}+d{V}^{-2}$$

(5)

4. Results and Discussion

4.1. Soaking Test for TiO₂

Soaking is a routine step for the electrolytic reduction of metal oxides and spent oxide fuels, which allows time for the system to reach stability and for salts to diffuse into the metal oxides matrix before the lithium metal is formed [11,27,28]. The soaked TiO₂ sample in LiCl–Li₂O (1 wt.%) at 650 °C without electrolysis consisted of three phases, namely, 60.8 wt.% of TiO₂, 38 wt.% of Li₂TiO₃, and 1.2 wt.% of Li_0.54Ti_2.86O₆, based on the XRD analysis (Figure 1). The Li₂TiO₃ must be formed chemically via the insertion of Li₂O into TiO₂ on the surface of TiO₂ particles through Equation (6), which is analogous to the formation of perovskite CaTiO₃ during the TiO₂ reduction in calcium salt melt systems [13,14,15,23]. However, the formation of Li₂TiO₃ is spontaneous in LiCl–Li₂O (1 wt.%) at 650 °C, while the formation of CaTiO₃ is slow and regarded as the rate-limiting step for the FFC process [6].

A total of 60.8 wt.% of TiO₂ remained after 5 h of soaking, which suggested that the diffusion of Li⁺ and O²⁻ ions through the increasing thickness of the Li₂TiO₃ layer had slowed down. The minor Li_0.54Ti_2.86O₆ phase also indicated that Li₂O non-stoichiometrically reacted with the bulk TiO₂ phase, possibly because only a very small portion of the Li₂O reached the fresh TiO₂ core due to the slow diffusion through the surface of the Li₂TiO₃ layer. This was the first time that the Li₂TiO₃ phase was structurally identified for the TiO₂ electrolytic reduction in LiCl salts, which was previously reported only based on the CV measurement of TiO₂ in molten LiCl (Table 1) [21].

4.2. Effect of the Reduction Potential on the Electrolytic TiO₂ Reduction

To understand the effect of the reduction potential on the TiO₂ electrolytic reduction, various cathodic potentials that were controlled directly using the reference electrode were applied for the TiO₂ reduction in LiCl–Li₂O (1 wt.%) at 650 °C using 80% of the theoretical charge. The lithium deposition potential was determined at −1.75 V vs. Ni/NiO, as discussed in a previous study [24]. Figure 2a shows the XRD results with Rietveld refinement analysis for TiO₂ electrolytic reductions at cathodic potentials of −0.2 V, −0.3 V, and −0.4 V vs. Li/Li⁺. For all three reduced samples, three phases were identified, including unreacted TiO₂, Li₂TiO₃, and LiTiO₂ compared with the standard XRDs (Figure 2a). No other reduced product or metallic Ti was detected. Lithium titanates, including LiTiO₂ and Li₂TiO₄, were reported for the TiO₂ electrolytic reduction in LiCl salts under various conditions, but no Li₂TiO₄ was formed in this study [10,19,20,21]. The quantitative compositional results based on the Rietveld refinement analysis for the three samples reduced at different cathodic potentials were compared with the soaked only TiO₂ sample (Figure 2b). Compared with the soaked only TiO₂ sample, the unreacted TiO₂ content at −0.2 V decreased from 60.8 to 14.5 wt.%, while Li₂TiO₃ remained at a similar level around 35.7 wt.%. The partial reduction product LiTiO₂ accounted for 49.8 wt.%. These results indicate that TiO₂ may have been directly reduced to LiTiO₂ by the in-situ-generated Li metal at −0.2 V vs. Li/Li⁺ (Equation (7)).

The TiO₂ content further decreased to 1.69 wt.% at −0.3 V, which lay under the XRD detection limit (~2%). The Li₂TiO₃ decreased from 35.7 to 28.3%, while the partial reduction product, LiTiO₂, increased from 49.8 to 70.4%. This suggested that both Li₂TiO₃ and TiO₂ were transformed into LiTiO₂ at a higher reduction potential (Equation (8)) and TiO₂ was more reducible than Li₂TiO₃. The reduction of Li₂TiO₃ to LiTiO₂ is a reversible reaction [29,30]. As the cathodic reduction potential further increased to −0.4 V vs. Li/Li⁺, LiTiO₂ stayed almost the same, while both TiO₂ and Li₂TiO₃ only changed slightly within the XRD limit.

Increasing the cathodic reduction potential could promote the TiO₂ electrolytic reduction to LiTiO₂ to a certain extent, then it started to reduce the current efficiency. A previous study also showed that a high Li metal concentration impeded the O²⁻ diffusion and led to a lower reduction extent of Ti⁴⁺ [31]. A similar trend was also observed in CaO/CaCl₂ for the TiO₂ electrolytic reduction [13]. In this study, the highest achieved TiO₂ reduction extent was 18.6% at −0.3 V vs. Li/Li⁺ using 80% of the theoretical charge (Figure 2b). The optimal reduction potential for TiO₂ reduction in the LiCl–Li₂O (1 wt.%) salt was determined as being −0.3 V vs. Li/Li⁺.

The I(t) curves for the TiO₂ reduction at all three cathodic reduction potentials showed a similar three-stage pattern, namely, a sharp current peak in the very beginning (<20 min), a current decrease, and a steady stage remaining between 0.4 and 0.6 A (Figure 3). The I(t) curves for the TiO₂ electrolytic reduction were significantly different from the I(t) curves for the NiO and Cr₂O₃ reductions, which showed high current plateaus for an extended time until the reduction finished [24,32,33]. The same interval operation mode was used in our previous NiO study to prevent the excessive accumulation of in-situ-generated Li metal [24]. The overall current magnitude increased when the reduction potential increased from −0.2 to −0.4 V (vs. Li/Li⁺), which decreased the experimental time for completing 80% of the theoretical charge for 2 g TiO₂ from 6.5 to 5 h. The FFC process of TiO₂ reduction also produced a similar I(t) profile [14,34]. The first stage of the current increase was attributed to the transformation of the insulator TiO₂ to lithium titanate LiTiO₂, which has a conductivity of 0.5 S cm⁻¹ [35]. According to the three-phase interlines (three PIs) reaction mechanism, the conductive LiTiO₂, insulator TiO₂, and electrolyte LiCl can form three PIs [36,37,38]. As the three PIs expanded from contacting points to a large area, the charge transfer reactions occurring at the three PIs led to an increasing current flow [36]. With more LiTiO₂ generated, the interline moved inward to the inside of the TiO₂ core. The longer diffusion distance for the three PIs resulted in a sharp current decrease at the beginning of the second stage (Figure 3). When the current continued to decrease at a slower rate, the Li₂TiO₃ shell started being reduced to the more conductive LiTiO₂ and three new PIs of LiTiO₂|Li₂TiO₃|electrolyte were formed. The continuous propagation of the three newly formed three PIs increased the material transfer distance, leading to the continuously decreasing current. The following stage varied little with time (around 0.4 A), indicating that either the reaction at the three PIs was limited by the slow diffusion of reactive species or it had reached equilibrium. For the TiO₂ electrolytic reduction at −0.2 V vs. Li/Li⁺, the final product still contained TiO₂, while the Li₂TiO₃ remained the same, suggesting that the final charge transfer was more relevant with reactions at the three PIs of LiTiO₂|TiO₂|electrolyte, where the O²⁻ diffusion coefficient was estimated as being 1.12 × 10⁻⁵ cm² s⁻¹. At −0.3 V and −0.4 V vs. Li/Li⁺, the reaction stopped at the reduction of Li₂TiO₃ at the three PIs of LiTiO₂|Li₂TiO₃|electrolyte, where the O²⁻ diffusion coefficients were estimated as being 1.46 × 10⁻⁶ cm² s⁻¹ and 1.68 × 10⁻⁵ cm² s⁻¹, respectively [21,36,37,38]. The I(t) curves of the TiO₂ reduction were a clear indication of the multistep reaction mechanism, which was limited by the O²⁻ diffusion from the three PIs to the bulk salts [13,15].

4.3. Effect of the Applied Charges on the TiO₂ Reductions at −0.3 V vs. Li/Li⁺

Using 80% of the theoretical charge only led to an 18.6% reduction of TiO₂. Thus, further investigation was carried out for TiO₂ electrolytic reductions by employing increased charges, namely 130% and 150% of the theoretical charge, at the optimal cathodic potential, namely, −0.3 V vs. Li/Li⁺.

The XRD analysis showed that no metallic Ti was formed for the TiO₂ electrolytic reduction when using 130% and 150% of the theoretical charge at −0.3 V vs. Li/Li⁺ (Figure 4a). The TiO₂ reduction extent increased from 22.3 to 25% when using 130% and 150% of the theoretical charge, respectively; meanwhile, the 13.4 wt.% of Li₂TiO₃ was completely converted to LiTiO₂ (Figure 4b). The 28.3 wt.% of Li₂TiO₃ remained under 80% of the theoretical charge and required 70% more of the theoretical charge to be reduced completely to LiTiO₂, suggesting the passivation behavior and stability of Li₂TiO₃. Furthermore, the formation of Li₂TiO₃ in the soaking stage may have slowed down the reduction process in several ways: (1) Li₂TiO₃ slowed down the outward diffusion of O²⁻ and (2) the reduction of Li₂TiO₃ was a reversible process, which could be controlled by the O²⁻ concentration [21,39].

These results show that the TiO₂ electrolytic reduction was very refractory, with LiTiO₂ being the most reduced product using 150% of the theoretical charge within 9 h, which was most likely due to the stability of Li₂TiO₃ and LiTiO₂. The sample geometry played an important role, as the adverse effect of Li₂TiO₃ and LiTiO₂ on the TiO₂ reduction is less pronounced in thin-film and fine powder TiO₂ [10,32]. The FFC process also shows that the electrolytic reduction of TiO₂ in CaCl₂ was challenging with a current efficiency that was typically less than 20%, although the much higher temperatures are more kinetically favorable for electrolytic reduction [40]. For the FFC process, the major products obtained within 8 h were a mixture of TiO, CaTiO₃, and CaTi₂O₄, and it took 120 h to completely reduce 8 g TiO₂ to Ti in CaO/CaCl₂ at 900 °C [14]. The electrochemical or chemical formation of various more thermodynamically stable perovskite phases as the intermediate products is one of the main problems contributing to the low current efficiency. The formation of lithium titanates is much easier and faster because the smaller Li⁺ ion leads to faster and deeper intercalation chemical reactions. However, LiTiO₂ is more thermodynamically stable than LiCl [40].

Similar I(t) profiles with a three-stage pattern are observed for TiO₂ reductions at −0.3 V vs. Li/Li⁺ using 130% and 150% of the theoretical charge but with an extended time (Figure 5). It took 8.2 h to apply 150% of the theoretical charge as the longest experiment among all the TiO₂ electrolytic tests in this study, with the highest reduction extent of 25% of 2 g TiO₂. The steady stage reflects the final charge transfer for the conversion of Li₂TiO₃ at the three PIs of LiTiO₂|Li₂TiO₃|electrolyte [21]. The O²⁻ diffusion coefficient in the steady state for 130% and 150% of the theoretical charge were estimated as being 8.98 × 10⁻⁷ cm² s⁻¹ and 4.35 × 10⁻⁷ cm² s⁻¹, which were about one order smaller than that for the steady state when using 80% of the theoretical charge. The smaller O²⁻ diffusion coefficients at higher theoretical charges indicates that the conversion of Li₂TiO₃ to LiTiO₂ was very refractory in the steady stage, which explains the extended reduction process, and hence, the time taken.

4.4. TiO₂ Reduction Mechanism

By employing a systematically controlled cathodic potential reduction and charges, reactions (6)–(8) were revealed and discussed for the TiO₂ electrolytic reduction process.

The DFT modeling was calculated based on the optimized bulk structures of the relevant compounds, including Li and Ti (Table S2). Gibbs free energy changes for the possible reactions involved during the TiO₂ reduction (Equations (6)–(13)) were calculated (

Table 2. Gibbs free energy change and reaction potentials for reactions at 650 °C using DFT calculations.

Reaction	Equation	∆Gr (kJ·mol−1)	Theoretical Potential (V)
${\mathrm{TiO}}_2+{\mathrm{Li}}_2\mathrm{O}\to {\mathrm{Li}}_2{\mathrm{TiO}}_3$	(6)	−125.0	--
$\mathrm{Li}+{\mathrm{TiO}}_2\to {\mathrm{LiTiO}}_2$	(7)	−113.0	--
${\mathrm{Li}}_2{\mathrm{TiO}}_3+\mathrm{Li}\leftrightarrow {\mathrm{LiTiO}}_2+{\mathrm{Li}}_2\mathrm{O}$	(8)	−11.0	--
${\mathrm{LiTiO}}_2+\mathrm{Li}\to \mathrm{TiO}+{\mathrm{Li}}_2\mathrm{O}$	(9)	0.8	−0.01
$\mathrm{TiO}+2\mathrm{Li}\to \mathrm{Ti}+{\mathrm{Li}}_2\mathrm{O}$	(10)	−22.3	--
$\mathrm{TiO}\to \mathrm{Ti}+1/2{\mathrm{O}}_2$	(11)	227.0	−1.18
${\mathrm{Li}}_2{\mathrm{TiO}}_3\to {\mathrm{LiTiO}}_2+\frac{1}{2}{\mathrm{Li}}_2\mathrm{O}+\frac{1}{4}{\mathrm{O}}_2$	(12)	222.0	−2.30
${\mathrm{LiTiO}}_2\to \mathrm{TiO}+\frac{1}{2}{\mathrm{Li}}_2\mathrm{O}+\frac{1}{4}{\mathrm{O}}_2$	(13)	212.0	−2.20

).

The negative Gibbs free energy ∆G_{650 °C}= −125 kJ mol⁻¹ for the insertion reaction of Li₂O to TiO₂ indicates that the formation of Li₂TiO₃ was spontaneous at 650 °C (Equation (6)), which is consistent with the TiO₂ soaking only testing result. The $∆G_{650 °C}$ for the TiO₂ reduction with Li metal was more negative than that for the reduction of Li₂TiO₃ with Li metal, suggesting that the reduction of TiO₂ with Li was more favorable than Li₂TiO₃ (Equation (7) vs. Equation (8)). The remaining core of the TiO₂ was most likely reduced to LiTiO₂ before the Li₂TiO₃ shell reduction, which could be slowed down by the Li₂TiO₃ passivation layer. The unit cell volume of Li₂TiO₃ was three times as much as the one for TiO₂, which not only reduced the pore volume for the transport of O²⁻ ions but was more thermodynamically stable (Figure 6) [6,40]. The reduction of Li₂TiO₃ by the Li metal generated Li₂O (Equation (8)), which needed to diffuse through the LiTiO₂ layer to the bulk molten salts. The slow diffusion of O²⁻ through the LiTiO₂ layer led to the accumulation of trapped Li₂O in the solid particle, which limited the reversible reduction of Li₂TiO₃. The further reduction of LiTiO₂ by the Li metal also generated Li₂O (Equation (9), ∆G_r = 0.8 kJ mol⁻¹). The direct electrolytic reductions of Li₂TiO₃ and LiTiO₂ were alternative pathways for the titanium dioxides’ reduction at underpotentials (Equations (12) and (13)), for which the Gibbs free energies were much higher. Both reduction pathways of LiTiO₂ were most likely halted due to the high local Li₂O concentration (Equations (9) and (13)), which became the key limiting step for further reduction and caused the extended time. The DFT analysis was consistent with the experimental TiO₂ electrolytic reduction results obtained in this study showing that the LiTiO₂ was the most reduced product achieved, even when using 150% of the theoretical charge. The passivation and refractory behaviors of LiTiO₃ were witnessed for the electrolytic reduction of TiO₂ in the LiCl molten salts using underpotentials before the formation of the Li metal [19,40].

By combining the experimental and DFT results, a three-step reaction mechanism for the TiO₂ electrolytic reduction at overpotentials in the LiCl–Li₂O (1 wt.%) salt at 650 °C was proposed for this study:

Step 1: $\mathrm{L}{\mathrm{i}}_2\mathrm{O}+\mathrm{T}\mathrm{i}{\mathrm{O}}_2\to \mathrm{L}{\mathrm{i}}_2\mathrm{T}\mathrm{i}{\mathrm{O}}_3$;

Step 2: $\mathrm{T}\mathrm{i}{\mathrm{O}}_2+\mathrm{L}\mathrm{i}\to \mathrm{L}\mathrm{i}\mathrm{T}\mathrm{i}{\mathrm{O}}_2$;

Step 3: $\mathrm{L}{\mathrm{i}}_2\mathrm{T}\mathrm{i}{\mathrm{O}}_3+\mathrm{L}\mathrm{i}\leftrightarrow \mathrm{L}\mathrm{i}\mathrm{T}\mathrm{i}{\mathrm{O}}_2+\mathrm{L}{\mathrm{i}}_2\mathrm{O}$ (reversible).

As illustrated in Figure 7, the soaking process followed a shrinking core mechanism, where the insertion of Li₂O and TiO₂ formed a stoichiometric Li₂TiO₃ layer on the external surface of the TiO₂ particles [16]. The Li₂TiO₃ layer became thicker and slowed down the O²⁻ diffusion, leading to unstoichiometric Li_0.54Ti_2.86O₆ formation with the unreacted TiO₂ core.

The Li₂TiO₃|TiO₂|electrolyte interline initially formed on the surface of the oxide particle and gradually moved to the interior of the particle.

In the second step, the Li metal diffused through the Li₂TiO₃ and reduced the TiO₂ core directly into LiTiO₂. The three PIs of LiTiO₂|TiO₂|electrolyte were formed. The lithium metal continuously moved to the center of the TiO₂ core until the core was fully converted to LiTiO₂. This step was controlled by the Li diffusion through the Li₂TiO₃ and newly formed LiTiO₂ layer. The second reaction also followed a shrinking-core model. When the TiO₂ was completely reduced to LiTiO₂, the Li₂TiO₃ shell started to be reduced to a mixture of LiTiO₂ and Li₂O, which was different from the inner LiTiO₂ core. The reaction occurred at the external layer of Li₂TiO₃ and formed a new three PIs of Li₂O+LiTiO₂|Li₂TiO₃| electrolyte. It stopped when the interline approached the LiTiO₂ core. In the I(t) curves, the second current reduction session took a long time and the data presented in Figure 4a shows that the conversion of Li₂TiO₃ to LiTiO₂ was not completed until 150% of the theoretical charge was applied. This indicates that Li₂TiO₃ was stable and the reduction of Li₂TiO₃ was refractory. Therefore, the third step was the rate-limiting step in the TiO₂ reduction in the molten LiCl–Li₂O.

Important physical variables determine the electrolytic reduction process, including the diffusion coefficient of oxygen in various phases, the applied current density, the radius of the grains, and the density of the phase [16]. In this study, the most reduced product obtained for the electrolytic reduction of TiO₂ was LiTiO₂ with 150% of the theoretical charge in 8.2 h. Previous studies on TiO₂ electrolytic reduction using the pellet form (thickness ≈ 0.2 cm) at cell potentials of 1.8 V and 3.2 V also only achieved LiTi₂O₄ and LiTiO₂ within several hours [19]. In contrast, the complete reduction of the TiO₂ powder in LiCl was achieved with both CV techniques and potentioamperometry [10,21]. Li₂O accumulation within the pellet may alternate the conductivity of the mixtures and slow down the direct electrolytic reduction of Li₂TiO₃ and LiTiO₂, which was observed for other metal oxide reductions [27].

The key to improving the TiO₂ reduction in LiCl–Li₂O molten salts is to reduce the Li₂O local concentration and promote the diffusion of O²⁻ out of the solid phase. Strategies to achieve this include lowering the Li₂O concentration in the molten salts, reducing the pellet size, and applying sonication or mechanical stirring to accelerate the O²⁻ diffusion [41]. Furthermore, as the Ti species valence reduced to 0 step by step according to the DFT analysis for the multistep reduction, the ∆G changed from negative to positive. This indicates that the stepwise reduction of TiO₂ with the in-situ-generated Li metal reached thermodynamic equilibrium and gradually became more refractory, although the overall ∆G for the TiO₂ reduction was negative and the reaction was spontaneous. Another promising way to improve the TiO₂ electrolytic reduction in LiCl is to avoid the intermediate steps, such as removing the soaking step or decrease it to a reasonably short time.

5. Conclusions

Electrolytic TiO₂ reduction in LiCl–Li₂O salts at 650 °C was examined in terms of a soaking only test, the effect of a reduction potential, and the effect of applied charges. The soaking without electrolysis showed the spontaneous formation of Li₂TiO₃, which is a unique intermediate for TiO₂ electrolytic reduction. The increasing reduction potential promoted TiO₂ reduction only to a certain extent. The optimal cathodic reduction potential was determined as being −0.3 V vs. Li/Li⁺. The highest charge, i.e., 150% of the theoretical charge, was applied for 8.2 h and caused the partial reduction of TiO₂ through to the full conversion of LiTiO₂, which was the most reduced product obtained in this study. Accordingly, the highest reduction extent was 25%.

Combined with the DFT modeling, the electrolytic reduction of TiO₂ in molten LiCl–Li₂O was proposed to follow a multistep reaction mechanism. The external surface of the TiO₂ particles was converted to Li₂TiO₃ via the chemical insertion of Li₂O during the soaking stage. Then, the TiO₂ core was partially reduced to LiTiO₂ before Li₂TiO₃ was electrolytically reduced to LiTiO₂. The reduction of Li₂TiO₃ was reversible and led to the formation of Li₂O. Due to the accumulation of Li₂O at the three PIs, the reduction of Li₂TiO₃ to LiTiO₂ was very slow and the rate-limiting step. Further reduction of LiTiO₂ was paused as more Li₂O was generated and accumulated.

Based on the understanding of the TiO₂ electrolytic reduction mechanism in molten LiCl–Li₂O, the key to increasing the efficiency of electrolytic TiO₂ reduction is to reduce the O²⁻ ions concentration at the three PIs. Strategies to achieve this include promoting the O²⁻ diffusion at the three PIs, using smaller TiO₂ particles or powder, using a lower concentration of Li₂O for the bulk salts, and reducing the soaking time to avoid the excessive formation of Li₂TiO₃.