3. Results and Discussion
In
Figure 1, the field emission scanning electron microscopy (FESEM) characterization of the samples T150 and T450 was reported. In
Figure 1a,b, the material composed by flakes, few nanometers thick, arranged in a porous 3D structure, is shown with two different magnifications. The observed material porosity results from the gas evolution during the oxidation process in hydrogen peroxide.
In
Figure 1c,d the FESEM characterization for sample T450 was reported at the same two magnifications. It is possible to observe a material with the same 3D arrangement as T150, but with the flakes composed by small crystals, interconnected along the flake wall. These crystals were the result of the calcination step operated in air at 450 °C.
To shed more light on the structural properties of the two materials, scanning transmission electron microscopy (STEM), high resolution TEM (HRTEM) and electron diffraction (ED) were performed.
Figure 2a shows the STEM-bright field image of a representative flake of sample T150, well transparent to the electrons, and quite homogeneous in thickness. HRTEM was performed on different flakes, to better understand the structural properties of such material. It resulted in being amorphous, with some very small (around 5 nm or less) crystalline domains present in the amorphous matrix, compatible with anatase crystalline structure. This was evidenced by the fast Fourier transform (FFT) in
Figure 2c, related to the white squared region in
Figure 2b, containing a small crystal, in which the dots represent the evidence of the crystalline orientation of the particle, and by the electron diffraction pattern of some flakes reported in
Figure 2d, constituted by rings (the amorphous part) and by dots (the crystalline part).
Figure 2e shows a HAADF-STEM image of sample T450; it represents a nanostructured material, composed by small crystals organized in flakes, as was already put in evidence from FESEM characterization. HRTEM in
Figure 2f shows crystals in the range 10–30 nm, identified as TiO
2 anatase. This is also confirmed by the electron diffraction pattern shown in
Figure 2g.
XPS analysis was carried out in order to study the surface chemical properties of both samples. From survey spectra (not reported) it was evident the presence of Ti, O, and C; the latter element was due to contamination related to ambient exposure from samples production to analysis session. We focused our attention on HR spectra for all the elements. Here we reported only Ti 2p signal (see
Figure 3), since it is the one that gave information regarding Ti average oxidation states. The Ti 2p
3/2 XPS core line spectrum for sample T150 (
Figure 3—left panel), was almost identical to that expected for a Ti(IV) chemical shift [
10], with a prominent peak located at 458.8 eV. However, by comparing and overlapping T150 and T450 experimental curves, a small shoulder was detectable only on the lower binding energy (BE) side of the signal spectrum for T150 sample, at 457.3 eV, which was the position usually related to Ti(III) oxidation state [
11], while this bump was not present in the T450 one. In order to evaluate the relative percentage due to the Ti(III) component, we fixed the deconvolution parameters by setting the FWHM of the Ti(IV) component equals to 1.1 eV and the Gaussian-Lorentzian ratio to 80:20 for both samples. For sample T450, only one component was enough to reproduce the experimental data, while for T150 one, an extra component was necessary to perfectly overlap the raw signal. This second component represents 2% of the entire signal. So, we could state that while for sample T450 we were able to assign only a Ti(IV) component to the Ti 2p signal, as expected for pure anatase phase, sample T150 surface was not completely uniform, from the structural and chemical point of view, since we detected two different phases: 98% Ti(IV) and 2% Ti(III).
Electrochemical properties of the two samples were studied by cyclic voltammetry of Li-insertion. In the T450 sample, the presence of strong peaks (called A peaks) at 2.00 V and 1.74 V, ascribable to the anatase phase, confirmed the phase purity of this sample. The voltammogram of T150 instead, exhibited two pair of peaks: A and D. Even though the structural properties of this sample are quite complex, it is possible to confirm the presence of small crystallites of anatase, rising the two small A peaks, in a matrix that produces stronger D peaks, as already evidenced in the TEM analyses. Those anodic and cathodic peaks that have their maximum intensities at 1.65 V and 1.35 V respectively show a large peak separation, as shown in
Figure 4, generated from an irreversible electrochemical reaction. The Li-insertion and extraction at those potentials produced extremely wide peaks, due to the absence of preferential channels for the intercalation that in turn provided different energy sites. These peaks represent a voltammetric signature that usually is observed in amorphous electrodes [
12,
13,
14]. Remarkably, the reduction of T150 covers a potential range wider than fully crystalline T450, and the sample delivered a substantial amount of charge at a potential below 1.7 V, suggesting a higher Li molar fraction with respect to that of pure anatase (x = 0.5) [
15,
16,
17].
The formal potential of D peaks, i.e., 1.50 V, is extremely close to the formal potentials of black TiO
2 (B) that are 1.52 V and 1.59 V [
18,
19], but the measurements recorded by cyclic voltammetry suggest a different nature: at rates lower than 1000 µV·s
−1, TiO
2 (B) usually exhibited two pair of peaks, S1 and S2, which were mainly capacitive. The D peaks in T150 instead were unique even at 50 µV·s
−1 and the charging and discharging processes appeared mainly diffusive. Not all the authors identify a capacitive charging in TiO
2 (B), in fact Mason [
20] observed a fully diffusion-controlled trend while Dylla [
21] and Laskova [
22] suggest a mixed capacitive and diffusion-controlled charging.
In order to address the complex behavior of the Li intercalation in the electrodes, the method of Dunn [
9] was employed: the peak current is analyzed at a fixed potential and obeys to:
where
kaν is the capacitive current contribution associated to the storage of Li
+ at the surface and to the bulk pseudo-capacitance, and
kb ν
1/2 is the diffusion controlled current, which is attributed to Li
+ insertion in the crystalline structure. It is clear from
Figure 4b, that a linear relationship was present only when the peak currents at the anodic and cathodic potentials (mentioned above for both the samples) were proportional to the square root of the scan rates. This indicates in the anode a prevalent diffusion-controlled charging and discharging through Li
+ in the crystalline lattice. Therefore, the diffusion coefficient was estimated by using the equation for an irreversible electrochemical reaction at a planar electrode:
where
ip is the peak current in amperes,
n and
na are taken as unity,
A is the BET inner surface area in cm
2,
C is the maximum concentration of Ti reduced from 4+ to 3+ in the lattice, in which x = 0.5 for the anatase phase; and
D is the chemical diffusion coefficient for Li
+ ions in cm
2/
s.α, the transfer coefficient, is estimated by applying the equation for irreversible waves:
where
na is the number of electrons involved in the rate determining step,
Ep is the peak potential, and
Ep/2 is the half-peak potential. The other symbols have their conventional meanings.
For the diffusion coefficient calculation, we also assumed that the thickness of reduced Ti3+ states is sufficiently thin with respect to the particles so that the particle surface can be considered planar.
From the slopes of
Figure 4b calculated by linear fitting and applying Equation (1), the diffusion coefficients have been estimated to be about 2 × 10
−17 cm
2/s in Li
+ extraction for T150 and 8 × 10
−16 cm
2/s and 3 × 10
−16 cm
2/s in extraction and insertion respectively, for T450, all at 1000 µV·s
−1. The diffusion coefficient in insertion of T150 was not estimated because of the impossibility to calculate the transfer coefficient in the reduction wave.
It is possible to notice that for both T150 and T450 samples, the Li
+ proceeds faster in extraction with respect to the insertion, as demonstrated by the slopes in
Figure 4b.
A better comprehension of the physical and electrochemical properties exhibited by the T150 and T450 samples can be obtained by examining the colorimetric properties. In fact, the oxidation of titanium foils in hydrogen peroxide produces samples, which—before annealing treatments—are extremely dark in color. In
Figure 4c these samples are indicated as “BA”, they clearly evidence the colorimetric behavior of the black titania (TiO
2 (B)). The BA samples have not been examined in this work since, for applications in electrodes of electrochemical storage devices, the peroxide molecules have to be removed by annealing treatments. After the annealing at 150 °C, the samples became lighter (ΔL* = +16), the yellowish disappears (Δh = +30), and the samples became duller (ΔC = −8), showing a deep grey color. After the second annealing at 450 °C, the lightness and chroma remained almost stable, but a massive change in hue was observed (Δh = −78). The variation of the interaction between the illuminant D65 and the samples at different annealing temperature were quantified and shown in the CIE L*Ch color space in
Figure 4c.
Due to the unique nature of T150 sample, in which an amorphous matrix contains small crystallites of anatase, the Dunn’s method was employed to distinguish the nature of the charging currents at each potential of the voltammograms. According to the method, herein we reconstructed the voltammograms (
Figure 5a) for the capacitive contribution (grey area) with respect to the raw voltammetric data of the overall currents (blue thick line). Therefore, the cyan area contribution reconstructed the diffusion-controlled charges.
The voltammograms at 50 and 400 µV·s
−1 are shown in
Figure 5a: it is clear that the capacitive contribution was remarkable only during the discharge of the electrode, confirming a larger diffusion constant during the extraction of Li
+.
Figure 5b also shows the percentage of the capacitive and diffusion-controlled processes to the overall charges collected at all scan rates recorded, either for the cathodic or anodic sweeps (the latters are patterned). The calculated contributions of capacitive and diffusion-controlled percentages with the extracted (Q
+) and inserted (Q
−) charges are reported in
Table 1. The pseudo-capacitive contribution in T150 dominated only for scan rates higher than 600 µV·s
−1, but at this scan rate the extracted charges were only half with respect to those measured at 50 µV·s
−1.
During the reduction of Ti
4+ to Ti
3+, the pseudo-capacitive contribution dominated only at scan rates higher than 1000 µV·s
−1 with a stored charge of only 30% with respect to that at 50 µV·s
−1. The reasons of this inequality during the insertion and extraction of Li
+ in this modified TiO
2 relied on the fact that the tetragonal structure of TiO
2 (poor in Li), containing distorted TiO
6 octahedra [
23], was much more favored and stable with respect to the orthorhombic structure (rich in Li); therefore the oxidation results tended to be faster and at lower overpotentials.
The partial capacitive nature of T150 could be either attributed to the presence of the anatase crystallites, which are nano-sized: for TiO
2 below 10 nm in size, it is well known that the capacitive contributions become relevant [
24], as shown in
Figure 2b. This approach was not employed to analyze the behavior at different scan rates of the sample T450 because the insertion and extraction of ions were fully diffusion-limited at the solid state, since the crystallites were larger than 10 nm, as previously observed in TEM.
To better investigate the potentialities/peculiarities of these materials, the self-discharge of T150 and T450 in a device-configuration was investigated. They were placed at their maximum charging potential (0.5 V) for 24 h and then let discharge at open circuit potential (OCP) by recording their drop in potential during the self-discharge. It is clear from
Figure 6a that the potential in T450 decayed linearly with respect to the square root of time, indicating a phenomenon in which the mass transport plays the crucial role and the diffusion from the bulk solution to the electrode leads the anode to discharge [
25]. This suggests the presence of some impurities at low concentration in the electrolyte, which are electro-active in the potential range 0.5 ÷ 2.8 V vs. Li/Li
+, and that shuttle the electrons from the working electrode to the counter electrode in which reduction occurs. Differently, in T150 the main process, which leads to the loss of charges at OCP, resulted in being under activation control. This evidence comes out from the linearity of the potential decay with respect to the Log of the time [
26].
It is well known that in amorphous materials, the cation mole fraction is higher with respect to the crystalline counterpart and this could generate a large concentration of highly reactive species leading to a faster discharge. These results should be taken into account when designing a full cell. In fact, the faster self-discharge of this electrode as anode could lead to a sudden return from the maximum potential of the full cell to 0 V.
To assess the stability of the two materials, aging cycling was performed. From the galvanostatic cycling of T450 in
Figure 6b it is possible to observe that in the first five cycles at 1C, the theoretical capacity of TiO
2 anatase (168 mAh·g
−1 [
27,
28]) was almost reached (155 mAh·g
−1) and that it decreased to become stable up to 200 cycles at a value of 125 mAh·g
−1. From 300 to 1200 cycles, the capacity decreased monotonically and then it stabilized for cycles up to 1300 with a capacity of 47 mAh·g
−1. The same trend was observed when cycled at 5C, where it started from 65 mAh·g
−1 and stabilized at the 700th cycle with a capacity of 25 mAh·g
−1. In T150 instead, the capacities were much less, being at 1C 65 mAh·g
−1 and at 5C 25 mAh·g
−1, but no fading was observed.
For both samples, it was observed that up to 1550 cycles, the coulombic efficiencies were much higher than 80% (
Figure 6b).
4. Conclusions
In this paper we presented an extensive study of the electrochemical properties of TiO2-based electrodes for energy-storage applications. The materials were prepared via a simple and scalable technique, i.e., by thermally oxidizing a titanium foil in hydrogen peroxide. This synthetic path led to the formation of a porous three-dimensional structure formed by few nanometer thick flakes, as proved by FESEM imaging.
Two different thermal treatments were carried out aiming at obtaining (i) a defective structure, by annealing the as-grown material in Ar atmosphere at 150 °C, or (ii) stoichiometric TiO2, by air annealing at 450 °C. The former structure resulted to be constituted by anatase nanocrystals embedded in an amorphous matrix, rich in Ti-reduced states corresponding to Ti (III). On the other hand, the latter was found to be formed by defect-free anatase crystals, with a mean size of tens of nanometers.
The Li-ion intercalation was investigated; as a result, the defective material exhibited a higher Li molar fraction with respect to the crystalline one, with a prevalent diffusion-controlled charging-discharging, mediated by Li+.
A faster extraction, with respect to insertion, of Li-ions was observed in both samples.
For this reason, the Dunn’s method was used to unravel the Li-ion intercalation processes, thus allowing us to reconstruct and separate the capacitive and diffusive contributions of the overall charges, giving a comparative quantification among them.
As a final result, it turned out that while the anatase sample successfully reached the theoretical capacity of TiO2, keeping it stable for 200 cycles, the defective material exhibited a lower capacity, however succeeding in maintaining it constant even after 1550 cycles. This characteristic makes this material promising in application requiring a higher stability, such as high power Li-ion batteries.