Hao Yang and
Jenq-Gong Duh*
Department of Materials Science and Engineering, National Tsing Hua University, Hsinchu, Taiwan. E-mail: jgd@mx.nthu.edu.tw
First published on 24th March 2016
Anatase nanoplates were successfully synthesized by a sol–gel method and exhibit superior electrochemical performance in LIBs and SIBs. Nanoplates with 10 nm particle sizes were produced by thermally decomposed titanium–terephthalate hybrid materials. In LIBs, the anatase nanoplates deliver a 100 mA h g−1 capacity at 10C, maintaining 150 mA h g−1 capacity at 5C for 500 cycles. Anatase with high crystallinity exhibits fairly good rate capability in SIBs, delivering 53 mA h g−1 capacity at 30C. Using ex situ X-ray diffractometry and X-ray photoelectron spectroscopy, anatase was discovered to trap Na, forming metallic Ti0 and amorphous sodium titanate in the chemical states of Ti3+ and Ti2+. The sodium titanate is structurally stable and electrochemically reversible with the redox couple of Ti4+/Ti3+. It is suggested that pseudocapacitance comprises most of the capacity in the first cycle. After the activation, anatase gradually stores more capacity by insertion. Therefore, it is demonstrated that anatase nanoplates with this promising insertive/extractive host structure are promising anode materials for LIBs and SIBs.
TiO2 has been recognized as a promising anode material for both LIBs and SIBs.8,9 Several types of TiO2 have been reported as anode materials for LIBs, such as anatase, brookite, rutile and TiO2(B). Anatase and TiO2(B) can deliver superior rate capability and cycle life owing to their open Li-ion diffusion channels.10–12 In anatase, the Li-ion insertion behavior is accompanied by a two-phase transformation from a tetragonal (space group I41/amd) to an orthorhombic structure (space group Imma). Li accommodation at the interstitial sites is expressed by the following equation:
TiO2 + 0.5Li+ + 0.5e− ↔ Li0.5TiO2 | (1) |
LiTiO2 (space group I41/amd) is only observed in nano-sized TiO2 and appears below 1.5 V vs. Li/Li+. Recently, it has been claimed that the [100] and [010] channels are the probable paths for Na-ion transportation.13,14 Na-ion is predicted to be stored in the interstitial sites, but is restricted to the lattice due to its large radius. Although several investigators have attempted to determine the Na-ion storage process in anatase, different results were revealed. Kim et al. and Oh et al. reported that Na-ion does not undergo a phase transformation but intercalates into the anatase lattice with Ti4+/Ti3+ redox reactions.14,15 Contrastingly, Wu et al. observed that Na-ion would distort anatase to an amorphous structure with metallic titanium formation. They discovered that about 41% sodium could reversibly diffuse out of anatase, indicating a strong insertion behavior in the first cycle.16 This observation was contrary to the results from González et al., who claimed a pseudocapacitive behavior based on NMR analysis.17 Thus, it is still critical to study the Na-ion storage process in anatase.
The construction of two-dimensional nanostructures is expected to be a promising method for enhancing the performance of SIBs, since a short diffusion path assists Na-ion to diffuse inside the particles.18,19 In this study, high and low crystallinity anatase nanoplates based on the concept of metal–organic frameworks (MOFs) were prepared. MOFs are highly porous materials which have wide applications, such as gas storage, catalysis and sensing.20 However, they are often prepared by a costly and time-consuming solvothermal process with toxic solvents (DMF, DEF, etc.), thus limiting their applications in industry.21 In this study, the as-synthesized anatase nanoplates from a facile aqueous sol–gel method exhibit uniform nanopores and particle sizes. These structures contribute to superior performance, especially rate capability, in LIBs and SIBs. Ex situ XPS analysis also reveals that the Na-ion storage process in the first cycle is dominated by a pseudocapacitive behavior. After the activation, the storage process will involve insertion behavior.
The low crystallinity anatase nanoplates were synthesized by filtering the undissolved terephthalic acid from the mixed solution. The filtered solution was heated at 90 °C for 2 h, and then the product was harvested via centrifugation.
Ex situ XRD and XPS. The cycled electrodes were identified by X-ray diffractometry (Rigaku, TTRAXIII, Cu Kα) at 50 kV and 300 mA in a high angle mode. High-resolution X-ray photoelectron spectroscopy (HR-XPS, PHI-5000 Versaprobe-II, ULVAC-PHI) was also used to derive the oxidation states of the cycled electrodes and was carried out using a monochromatic Al kα source. The analyzed area on the sample was about 500 × 500 μm and was etched by C60+ sputtering at 20 kV 20 nA for 8 min. To perfectly fit the spectra for mixed oxidation states, some restrictions should be followed. The fitting parameters were edited from the NIST database using the C 1s signal (284.5 eV) as the internal standard. The oxidation states of Ti4+, Ti3+, Ti2+ and Ti0 were fixed at 458.80, 457.27, 455.48 and 454.00 eV, respectively. The integrated area of Ti 2p1/2 to 2p3/2 was in the ratio of 1:
2 and was split into 5.66, 5.60, 5.73 and 6.13 eV. It was noted that the FWHM value of each fitting spectrum was highly dependent on the crystallinity. The structure in disorder was slightly broader than that in order. In this study, anatase gradually transformed to an amorphous structure during discharge. Thus, all the fitting spectra were constrained to have identical widths that were broader than the fitting spectrum of the as-synthesized electrode.
In TEM images (Fig. 2a), AN shows some anatase clusters distributed in the amorphous structure, which is coincident with the XRD results. As is evident in Fig. 2b, the structure of AN-500 is around 50 nm in thickness; however, the widths are hundreds of nanometers to micrometers. The primary particles of AN-500 are around 10 nm and are connected to each other with uniform arrays (Fig. 2c and d). To identify the whole microstructures, the anatase nanoplates were analyzed by N2 adsorption–desorption isotherms, as shown in Fig. 2e. The specific surface area of AN (254.9 m2 g−1) is greater than that of AN-500 (91.6 m2 g−1). The pore size distributions derived from the Barrett–Joyner–Halenda method reveal that AN and AN-500 have average pore sizes of 2.93 and 3.69 nm, respectively. The narrow pore size distributions and small particle sizes demonstrate that terephthalic acid successfully restrains the aggregation of anatase nanoparticles, retaining the nanopores instead of destroying them.
![]() | ||
Fig. 2 TEM images of (a) AN and (b)–(d) AN-500. (e) The pore size distribution calculated from the BJH method. The inset shows the N2 adsorption–desorption isotherms of AN and AN-500. |
AN and AN-500 were tested as anodes in LIBs and SIBs using the corresponding metals as counter electrodes. In LIBs, both materials delivered higher rate capability than the commercial p25 (Fig. 3a). In the discharge curves (Fig. 3b), AN-500 shows distinct phase transformation plateaus at 1.72 V, contributing more capacity (90 mA h g−1) than AN (30 mA h g−1) at 0.5C. Although AN does not exhibit distinct phase transformation plateaus due to its low crystallinity, it delivers rather high capacity above 5C. Most of the capacity comes from the slope between 1.72 and 1 V. A slope without significant capacity fading is associated with pseudocapacitive behavior (as will be discussed later). This indicates that about 70% of the capacity (136 mA h g−1) comes from Li-ion that is stored near the surface due to the higher surface area. However, the large surface area also results in 190 mA h g−1 capacity loss in the first cycle, greater than that of AN-500 (60 mA h g−1).
In SIBs, the cells were operated at the rates of 0.5 and 0.05C. The voltage was driven to 0.01 V to promote more Na-ion diffusing into TiO2 (Fig. S1†). In the first cycle (Fig. 3d), AN shows oblique voltage curves at both 0.5 and 0.05C, exhibiting a typical solid–solution reaction.9 In contrast, AN-500 delivers a voltage plateau at 0.5C, and an additional plateau appears between 0.80 and 0.60 V as the rate decreases to 0.05C. At the beginning of sodiation, the voltage quickly drops from 2.60 to 0.65 V, contributing 65 mA h g−1 capacity. These voltage drops, which are independent of the current rate, are associated with the solid–solution reaction and SEI formation with electrolyte decomposition, independent of the insertion of Na-ion into the anatase lattice.14,23 The two subsequent plateaus located at 0.65 and 0.27 V are greatly elongated at 0.05C. Although there are trace amounts of S in AN-500, it shows no electrochemical behavior (Fig. S2†).24,25 It is interesting to note that AN-500 reveals similar coulombic efficiencies of 51.3% and 50.5% at 0.5 and 0.05C, respectively, indicating that the plateaus do not involve electrolyte decomposition, which will enhance the portion of irreversible capacity. Thus, the plateaus are associated with the reaction of phase transformation. However, it was also discovered that the lower plateau may involve additional formation of dendrites, which will be discussed later.
In rate capability tests (Fig. 3e), AN-500 shows comparable capacity to LIBs, delivering 200, 174, 153, 134, 104, 77, and 53 mA h g−1 at the rates of 0.5, 1, 2, 5, 10, 20, and 30C, respectively. In contrast to lithiation/delithiation, Na-ion shows no distinct phase transformation plateau (Fig. 3f). It is logical that the larger Na-ion radius (1.02 Å) cannot easily diffuse into the anatase lattice as Li-ion (0.76 Å) does.26 Na-ion is preferentially stored on the surface rather than in the lattice. The oblique voltage curves in Fig. 3f should be identified as a pseudocapacitor rather than a double-layer capacitor, which delivers a sloping voltage drop.11,27 Pseudocapacitors, such as TiO2, CeO2 and Nb2O5, are generally metal oxides with high specific surface areas that can exhibit high rate capability.28,29 However, in Fig. 3g, AN shows a low rate capability, even lower than that of commercial p25, despite having a large specific surface area. Thus, in comparing two galvanostatic voltage profiles (Fig. 3f and g); although AN shows high polarization with increasing current rate, the curves have similar shapes to those of AN-500, indicating a similar electrochemical mechanism. The same tendency is also observed in the CV:AN and AN-500 exhibit cathodic/anodic peaks after sufficient cycling. Therefore, any investigation into the poor rate capability of AN should focus on the first cycle.
To examine the rate limitation in AN, the CV at 0.2 mV s−1 for 10 cycles was tested, recording the Nyquist plot at the end of each cycle, as shown in Fig. 4. In the first cycle of AN, high concentrations of Na are trapped in TiO2 defects, leading to a huge cathodic peak at 0.26 V. On the other hand, AN-500 shows two cathodic peaks at 0.35 and 0.12 V, which are identified as the two plateaus in Fig. 3d. The relatively low voltages compared to the discharge curves are attributed to polarization at the high current rate (0.2 mV s−1 ≈ 0.24C). Following the irreversible cathodic peaks, a slight anodic peak located at 0.10 V represents the desodiation of super-P, which shows no electrochemical reaction above 1 mV s−1.30 It is evident that neither sample has a reversible redox couple in the first cycle. The couple gradually appearing at 0.60 and 0.84 V is recognized as the activation.31 Therefore, EIS is operated at 3 V to realize the cell resistance. In the Nyquist plot (Fig. 4c and d), the curves can be divided into sections of Rs, Rct and the Warburg coefficient. Rs is caused by the resistance of the cell, electrolyte and SEI. The semi-arc, which represents Rct, is associated with the charge-transfer resistance between the interfaces. In this case, the interface is equal to the region between the TiO2 surface and the electrolyte. After the first cycle, the Rct of AN-500 decreases and remains at 350 ohm, forming stable interfaces on the surface. In contrast, AN shows a gradual increase in Rct. It is demonstrated that Na is trapped by surface defects and residual organics (Fig. S3†), restraining the subsequent sodiation in AN.32
![]() | ||
Fig. 4 Cyclic voltammetry of (a) AN and (b) AN-500 at 0.2 mV s−1 in SIBs; AC-impedance of (c) AN and (d) AN-500 at the end of each cycle of cyclic voltammetry. |
To understand the phase transformation of AN-500, ex situ XRD at different states of charge in the first cycle was carried out, as shown in Fig. 5a and b. The electrode was cycled between 3 and 0.01 V at 0.05C. When the cell was discharged to 0.45 V, the diffractions broadened and (101) slightly shifted to lower angles compared to the as-synthesized electrode. It is thus confirmed that the plateau at 0.65 V belongs to the phase transformation from crystalline to amorphous structure. Following the plateau at 0.27 V, the XRD pattern at 0.01 V is totally amorphous, without any anatase diffraction. This disappearance of the diffraction is attributed to the insertion of Na-ion into the anatase lattice, which is accompanied by an anisotropic structural distortion.33 We also discovered that anatase retained poor crystallinity at 0.5C, implying that the length of the plateau was associated with the degree of structural distortion. However, AN-500 still retains an amorphous structure as it is charged to 3 V, demonstrating that the irreversible Na is still trapped by the lattice. Ex situ SEM images reveal both the cross-section and top view of the electrode (Fig. 5c–f and S4†). It is interesting that as it is discharged to 0.01 V, dendrites cover the top of electrode. These dendrites are formed below 0.45 V and dissolve above 0.9 V in the first cycle. After that, the dendrites experience high polarization and do not reappear in the next cycle. According to the EDX mapping (Fig. S5†), the dendrites are present in the formula of sodium superoxide (NaO2). However, even low-angle XRD cannot detect any diffraction related to the dendrites, which seem to be amorphous. It is thus difficult to clarify whether NaO2 was formed upon cycling or was oxidized by air when the cell was opened.34 To the best of our knowledge, the dendrite structures initially form in crystalline Na and are then destroyed by the air. This gives a reasonable explanation as to why the dendrites have regular shapes.35,36
Ex situ XPS was further explored for the cycled electrodes to reveal the oxidation states of anatase. In the literature, it was observed that Ti4+ would be reduced to Ti3+ and metallic Ti0 during sodiation.14–16 Wu et al. removed the SEI on the electrode by Ar+ sputtering and quantified a large insertion capacity within the first cycle.16 However, Ar+ is a strong reductive source that will highly destroy the structure as well as reduce Ti4+. This demonstrates that Ar+ sputtering is not suitable for quantifying different Ti oxidation states. The effect of sputtering is also revealed in Fig. S6.† Therefore, C60+ was chosen instead of Ar+ to sputter the electrode surface. Additionally, the accuracy of curve-fitting is another critical point for spectra involving several oxidation states. The fitting principles have been described in the Experimental section. The Ti spectra and the calculated results are shown in Fig. 5g and Table 1. It is revealed that as it is discharged to 0.45 V, Ti4+ is reduced to Ti3+ and Ti2+. The formation of Ti2+ is reported for the first time in this study. If it is assumed that both oxidation states are reduced by Na-ion, forming NaTiO2 and Na2TiO2, the total capacity calculated from XPS is consistent with the discharge capacity. Ti2+ comprises a 0.75 portion of the capacity at 0.45 V and does not increase at lower voltage, implying that the side effect of the solid–solution reaction above 0.45 V involves strong surface reduction. As the charge reached 0.01 V, a further 0.12 portion of Ti3+ formed, followed by a 0.06 portion of Ti0. The conversion of TiO2 is reported, yet the value in this study is not as high as that in the literature.16 It was revealed that the total capacity calculated from XPS is 217.1 mA h g−1, incompatible with the discharge capacity of 622.0 mA h g−1. The only explanation is the formation of dendrites at low voltage. As they are charged to 3 V, Ti2+ and Ti0 are fully electrochemically irreversible. Only a 0.10 portion of Ti3+ re-oxidizes to Ti4+, leading to 33.4 mA h g−1 capacity. The difference in charge capacity between XPS and the galvanostatic voltage profiles gives rise to an assumption that, in the first cycle, most of the reversible Na is stored on the TiO2 surface, forming the pseudocapacitance. When the cell is opened, Na on the TiO2 surface can be washed out by PC. Thus, Ti3+ without Na is re-oxidized by the atmosphere.
Oxidation states | Ti4+ (458.80 eV) | Ti3+ (457.27 eV) | Ti2+ (455.48 eV) | Ti0 (454.00 eV) | Total capacity (ex situ XPS) | Total capacity (galvanostatic voltage profiles of 0.05C) |
---|---|---|---|---|---|---|
a The chemical state was calculated from the pristine electrode sputtered with C60+ for 12 min.b The discharge capacity (mA h g−1) was calculated from the portion of Ti oxidation states and has been already subtracted the portion of Ti reduction from C60+.c The charge capacity (mA h g−1) was calculated from the portion of Ti at higher voltage to subtract the portion at 0.01 V. | ||||||
Before cyclea | 0.91 | 0.07 | 0.02 | — | — | — |
0.45 V | 0.58 | 0.20 (43.4)b | 0.22 (133.6) | — | 177.0 | 180.0 |
0.01 V | 0.40 | 0.32 (83.5) | 0.22 (133.6) | 0.06 (80.1) | 217.1 | 622.0 |
0.9 V | 0.50 | 0.25 (23.4)c | 0.21 (−) | 0.04 (−) | 23.4 | 129.2 |
3 V | 0.52 | 0.22 (33.4) | 0.21 (−) | 0.05 (−) | 33.4 | 315.0 |
To understand the particularity of the anatase rate capability, AN-500 was cycled at several scan rates (Fig. 6a and c) which are described as a power-law relationship with redox peaks. The equation follows:
i(redox peaks) = aνb | (2) |
i(ν) = ia + ib = C1ν + C2ν0.5 | (3) |
Finally, AN and AN-500 were tested for long cycling stability. In LIBs (Fig. 7a), AN and AN-500 show comparable performance up to 500 cycles, maintaining coulombic efficiencies near 99%. AN shows activation in the first 100 cycles that may be associated with the rearrangement of the amorphous structure during lithiation/delithiation.9 When the SIBs were cycled at 0.5C (Fig. 7b), despite the poor rate capability delivered by AN, it still maintained 143 mA h g−1 capacity with 97% coulombic efficiency after 150 cycles. On the other hand, AN-500 shows good cycling stability within 150 cycles, maintaining 230 mA h g−1 capacity without decay. A comparison of this work with the literature is also represented in Table. S1.† In this study, the anatase nanoplates with 10 nm particle size exhibit greater capacity at low current. Among all the results, the anatase nanoplates are a promising anode which exhibits good cycling stability in LIBs and SIBs. However, operating the cell near 0 V to enhance the capacity causes a large amount of irreversible Na formation, forming the irreversible Ti3+, Ti2+, Ti0, and the Na-containing dendrites. Although the formation mechanism of the dendrites is still unclear, it is fortunate that they disappear at the end of the first cycle and do not reappear. In the aspect of a full battery, the supply of Na is limited by the cathode materials. Therefore, it is suggested that surface modification might be a suitable way to eliminate the dendrites, improving the coulombic efficiency of anatase (50.5%) in the first cycle.
![]() | ||
Fig. 7 Long cycling tests of AN and AN-500: (a) cycling at the rate of 5C for 500 cycles in LIBs, (b) cycling at the rate of 0.5C for 150 cycles in SIBs. |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra27814g |
This journal is © The Royal Society of Chemistry 2016 |