2.1. Galvanostatic Tests
The electrodes were fabricated using the “slurry” technology. For this purpose, titanium oxyfluoride powders were rubbed in an agate mortar and mixed with dry components in the weight ratio T1 (or T2): carbon black: VGCF (vapor grown carbon fibers, Showa Denko) = 88%: 5%: 2%, to which a binder solution consisting of N-methylpyrrolidone and polyvinylidene fluoride (PVDF) Solef 5130 (Solvay, Shanghai, China) was then added in a 17:1 ratio. Electrodes 15 mm in diameter were mounted in CR2032 disc cells with a lithium counter electrode and Celgard 2325 as a separator. The electrolyte used was TC-E918 (Tinci), which is a 1M solution of LiPF6 in an EC/PC/DEC/EMC/PA mixture (ethylene carbonate, propylene carbonate, diethyl carbonate, ethyl methyl carbonate, propyl acetate).
Charge–discharge curves of T1-Li (or T2-Li) half-cells were recorded in galvanostatic mode using a CT3008W-5V10mA (Neware, Shenzhen, China) cycle tester. During Li embedding (charging), the voltage across the test electrodes was limited to 10 mV, and during Li extraction (discharging), it was limited to
U = 3 V. The long-term tests, the results of which are shown in
Figure 1,
Figure 2,
Figure 3 and
Figure 4a, started with the first three cycles (
N = 1–3) of charge–discharge being carried out at current density
j = 30 mA/g, then three cycles (
N = 4–6) at
j = 100 mA/g, and another three cycles (
N = 7–9) at
j = 200 mA/g. Further tests were performed at
j = 100 mA/g. Cycling at
j = 400 mA/g, presented in
Figure 4b, followed the initial 12 cycles at
j = 30–800 mA/g (see
Figure 5a). The cycling was carried out without the cell being thermostatted.
The charge curve for the first lithiation is fundamentally different from subsequent cycles (
Figure 1). It has a long horizontal plateau which is characteristic of the materials forming two phases during lithiation. The position of this plateau is slightly different for the two materials: in the case of T1, it is higher with an average value of 0.82 V, and in the case of T2, with an average value of 0.74 V.
Table 1 makes it possible to estimate the number of lithium g-atoms embedded in the different sections of the curve. The formula
x =
Q·
M/
F = 3.8 × 10
−3·
Q, where
M is the molar mass of TiOF
2 = 101.88 g/mol,
Q in mAh/g, and
F = 26,801 mAh/mol, was used for the calculation of lithium g-atoms number
x per TiOF
2 mole.
The structure change in TiOF
2 as a result of electrochemical lithiation at its different stages was studied in [
3,
5] using X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), and X-ray absorption near-edge structure (XANES). According to
Table 1, 0.16 and 0.24 Li atoms, respectively, are first introduced into the T1 and T2 material at voltages above the plateau. Under this stage, the structure of the material remains crystalline.
In the plateau region (range 2–3), the introduction of ~2 lithium g-atoms occurs. It is accompanied by the formation of LiF, partial reduction of Ti, and amorphization of the material [
3,
5]. The last lithiation range 3–4 corresponds to the introduction of Li into the amorphous phase. The maximum amount of lithium introduced during the first lithiation is 2.85 and 2.67 g-atoms for T1 and T2 materials, respectively. It can be assumed that the first lithium introduction results in a conversion reaction resulting in the formation of several compounds. It may correspond to an irreversible reaction involving 2 Li ions:
where the oxidation state of titanium changes from 4+ to 3+. Further lithiation of titanium compounds can continue with the participation of 3 Li ions, but this time reversibly:
which suggests two independent reactions:
The total reaction (1) and (2), corresponding to the first introduction of lithium, can be represented as:
Thus, 1 mole of titanium oxyfluoride in our supposed scheme of sequential lithiation absorbs 5/2 = 2.5 lithium atom, which is in good agreement with the data in
Table 1. In this case, lithium fluoride is considered to be inert, and the formation of free titanium (Ti
0) does not occur, which corresponds to the data of [
3,
5]. The irreversible reaction (conversion reaction) during the first introduction of lithium into TiOF
2 strongly reduces the Coulomb efficiency (CE) of the first cycle. For T1 and T2 materials, it is 63% and 60%, respectively. According to reactions (1)–(5), the irreversible losses of the first cycle are 2/5 = 0.4, which agrees well with the experimental values of CE. Thus, the electrochemical characteristics of both materials are extremely similar, differing only in the voltage at which the plateau appears and the value of the introduced/removed charge.
Figure 1b shows the S-curves of cycle 2, the characteristic points of these curves are given in
Table 2. Starting from the second cycle a reversible redox reaction identical to the second stage of the first lithiation (2) takes place. During the lithiation, titanium is reduced to Ti
2+ in TiO (3) and to Ti
+ in Li
2TiF
3 (4), which corresponds to an average valence value of Ti
1.5+. Hence, there are 3 Li atoms per 2 Ti atoms, i.e., Li/Ti ratio
x = 1.5, which corresponds to a theoretical capacity of 395 mAh/g. This value is slightly lower than the total absorbed lithium
x = 1.54 for T2 material (see
Table 2) found from experiments. For T1, this value is even greater:
x = 1.72, which can be explained by a higher content of TiF
3. TiF
3 can reversibly absorb 3 lithium atoms (its theoretical capacity 766.6 mAh/g) [
1]. Thus, at the end of the Li embedding process, titanium is reduced to Ti
1.5+ on average, and by the end of lithium extraction, it is oxidized to Ti
3+. The latter is in agreement with the data of [
3], which holds that the charge state of Ti decreases to values <3 during reversible lithiation, but differs from the conclusions of [
5], according to which Ti
2+/4+ reduction/oxidation occurs during lithium embedding/extraction. Since in our experiments for both materials the reversible capacity
Qexp was higher than that calculated for TiOF
2 (within the proposed model), this discrepancy allows us to estimate the amount of TiF
3 in each material using a simple equation:
where
y is the mass fraction of TiF
3. Such an estimation gives for T1 material
y = 15.9%, and for T2,
y = 2.7%.
2.2. Cycle Life
After the first nine cycles, the anodes T1 and T2 were tested at a constant current of
j = 100 mA/g.
Figure 2 shows the dependence of the discharge capacity on the number of charge–discharge cycles
N. The observed fluctuations are due to temperature changes in the room where the tests were carried out. After ~100–140 cycles, the lithium counter-electrode degrades, which appears as a capacity drop. Replacing the lithium in the cell returns the
Qch and
Qdch to their previous values. Fitting of the curves with linear dependence is expressed using the formulas
Qdch (T1) = 406 − 0.344
N and
Qdch (T2) = 350 − 0.051
N, i.e., the specific capacity of T1 material containing TiF
3 is higher than of T2, but the degradation rate is noticeably higher, too.
Figure 3 shows how the charge–discharge curves change from cycle 12 to cycle 80 for both materials. The changes are small and are most likely due to the increase in the electrical resistance of the materials.
The Coulomb efficiency behaves somewhat unusually. From
Figure 1b and
Table 2, it can be seen that in the second cycle, the amount of electricity extracted exceeds the amount injected by 0.26 and 0.51% for the T1 and T2 materials, respectively, and the Coulomb efficiency appears to be >100%.
Figure 4a shows the CE fluctuations occurring after the 12th cycle up to
N = 90 when tested at constant charge/discharge current value
j = 100 mA/g, and in
Figure 4b, at
j = 400 mA/g. Approximating the dependencies with a linear function and averaging over the samples indicates a slowly decreasing dependence with a slope of 0.002% per cycle for
j = 100 mA/g and a slope of 0.004% for
j = 400 mA/g. The crossing of the 100% boundary occurs between 40 and 60 cycles. It can be assumed that in this region, the degradation starts to override the effect of additional lithium release. The behavior of the Coulomb efficiency during long cycling in the galvanostatic regime seems to be a result of the fact that one of the lithium compounds formed in the conversion reaction during the first introduction of lithium (in this case LiF) does not remain completely inert and turns out to be able to contribute to the lithium yield during the reverse electrochemical reaction. The electrochemical decomposition of LiF and Li
2O is known to be activated with the participation of transition metals or their oxides [
14]. The gradual depletion of an additional source of Li occurs slowly but serves to cause the observed phenomenon where the observed CE exceeds 100% by 0.1–0.2% over several tens of cycles.
2.4. Cyclic Voltammetry (CVA)
The measurements were carried out on a Biologic VSP modular potentiostat/galvanostat. Experiment E1 used samples that had previously undergone two cycles. In this way, the third cycle dependencies were recorded when the voltages were sweeping at different speeds (0.05–2 mV/s). In experiment E2, the sweep rate was
v = 0.2 mV/s starting from cycle 1. In experiments E3 and E4, both materials were investigated, and the sweep speed
v = 0.05 mV/s. These experiments make it possible to compare the electrochemical properties of T1 and T2 materials and also to determine Li diffusion coefficients in each material.
Table 3 shows the details of the experiments and the parameters of the samples. The surface area
S of the samples was calculated from the average particle size
a in the starting material
S = 6
m/(
ρ·a); in both cases
a = 1 μm (SEM data in figure in
Section 3.2, SEM images).
One cathodic C and one anodic A peak are well pronounced. Their position is consistent with the flattest parts of the charge–discharge curves (see
Figure 1b and
Figure 3,
Table 2). As the sweep speed increases, the peaks move apart in voltage from each other and become blurred.
Figure 7 shows the dependencies of their respective voltage on
v. The position of the peaks for
v = 0 mV/s in
Table 4 is obtained using extrapolation of the curves. It can be seen that the dependences are nonlinear, which makes it impossible to explain the shift solely by the contribution of series ohmic resistance. The cathode peak is more sensitive to the sweep speed than the anode peak: the former is shifted in voltage by 0.41 V and the latter by 0.25 V. This result indicates the asymmetry of the kinetic properties during charging and discharging.
Figure 8 shows the dependence of the height of the anode (A) and cathode (C) peaks on the square root of the voltage sweep rate. Its linearity indicates that the main process limiting the charge and discharge rate is diffusion [
15].
Figure 9 shows the results of experiment two—the dependence on the number of cycles. In
Figure 9b, from the intermediate cycles between the second and eleventh one, it can be seen that as the number of cycles increases, the relative height of peaks 0.88 and 1.03 is changed. This corresponds to a decrease in height of the cathode peak 1.03 with its simultaneous shift towards higher voltages so that by cycle 5, it looks like a small shoulder. At the same time, the 0.88 cathode peak and the 2 V anode peak just slightly change their voltage position. With an increasing number of cycles, an additional 0.6 V anode peak and 2.02 V cathode peak appear.
Figure 10 shows the CVA of cycle 1 and cycle 2 recorded at low speed (experiments E3 and E4), and
Table 5 shows the parameters of the observed peaks. The difference between the T1 and T2 materials is present only in the first cycle where the 1C cathode peak for T1 is located at a higher voltage than for T2, the same as the plateau in the first charging curve. The presence of two peaks on the cathodic part of the curve for cycle 2 can testify to the Li introduction mechanism corresponding to the formation of different compounds Li
2O and Li
2TiF
3, (see Equations (3) and (4)). Both T1 and T2 materials exhibit significant hysteresis, which is clearly visible from the difference in voltages for the anodic and cathodic peaks (
Figure 10b) and from the voltage location of the gentle sections 2–3 and 5–6 on the charge and discharge curves (
Figure 1b and
Table 2). In addition to the polarization component, its presence is most likely due to the so-called path hysteresis [
16].
2.5. Determination of the Diffusion Coefficient
The mean value of diffusion coefficient
D of lithium in the electrode material was determined using the height of CVA peaks of the dependences recorded at low speed in experiments E3 and E4 separately for the cathode and anode half-cycle. For this purpose, the Randles–Shevcik equation was applied [
17,
18]:
where
c0 is the maximum concentration of lithium embedded in or extracted from the electrode in mol/cm
3,
S is the inner surface area [cm
2],
v is the voltage sweep rate [V/s],
Ip is the peak current [A]. The dimension of the coefficient is 2.69
× 10
5 [A·s·mol
−1·V
−1], and
z is an empirical fitting parameter. Equation (7) was derived from a modified Nernst equation [
15] defining the relation between the electrode potential and concentration of potential-determining particles and taking into account the difference between the concentration and activity of diffusing particles:
The necessity to introduce the fitting parameter
z is due to the fact that the equilibrium dependence of potential
E on
c in the general case does not obey the Nernst equation. At room temperature,
RT/
F = 2.57
× 10
−2 V, number of electrons
n = 1, and therefore (8) appears as:
Thus, comparing the slope
s of the empirical relationship
E = f(ln
c) with the slope of relationship (9) we can determine
z = 2.57
× 10
−2/|
s|. The relation between the lithium concentration [mol/cm
3] and the amount of electricity
Q [C] present in the electrode is determined using the formula:
where
ρ [g/cm
3] and
m [g] are the density and mass of the active component of the electrode, respectively, and the Faraday constant
F = 9.65
× 10
4 C/mol.
Let us first determine
D for the T1 material from CVA cycle 2 based on the height of cathode peak 3C
Ip = −0.684 mA (see
Table 5). To find
c0, we used the maximum charge accumulated at the cathode half-cycle,
QmaxC = 12.94 C, which was obtained by integrating the dependence
I = f(
t). The corresponding value was
c0 = 0.050 mol/cm
3. From the slope of the straight line in
Figure 11a, we determined
z = 0.066. Taking into account electrode surface area
S = 161 cm
2 from (7), we find
D = 3.02
× 10
−14 cm
2/s.
By integrating
I = f(
t) in the anode part of the CVA, it was found that
QmaxA = 12.43 C,
c0 = 0.048 mol/cm
3. The slope
E = f(ln
c)
s = −0.331, whence
z = 0.078 and
D = 1.24
× 10
−14 cm
2/s. A similar procedure for the T2 material (
Figure 12) made it possible to find the correction factor
z and to calculate the diffusion coefficients for the cathode and anode peaks.
Table 6 shows that the diffusion coefficients found for both materials are almost the same.
The obtained
D values are rather low. The low Li diffusion coefficient leads to a high polarization, i.e., a large deviation in the charge and discharge dependences from the equilibrium curves. As a result, Li insertion takes place at lower potentials relative to lithium, and extraction, respectively, at higher ones. Thus, with an increase in the current, a significant reduction in capacity occurs both for Li insertion and extraction. The effect of current on the discharge capacity can be seen in
Figure 5.