Effect of Hf-doping on electrochemical performance of anatase TiO2 as an anode material for lithium storage

Hafnium-doped titania (Hf/Ti = 0.01; 0.03; 0.05) had been facilely synthesized via a template sol–gel method on carbon fibre. Physico-chemical properties of the as-synthesized materials were characterized by X-ray diffraction, Raman spectroscopy, scanning electron microscopy, energy-dispersive X-ray analysis, scanning transmission electron microscopy, X-ray photoelectron spectroscopy, thermogravimetry analysis and Brunauer–Emmett–Teller measurements. It was confirmed that Hf4+ substitute in the Ti4+ sites, forming Ti1–xHfxO2 (x = 0.01; 0.03; 0.05) solid solutions with an anatase crystal structure. The Ti1–xHfxO2 materials are hollow microtubes (length of 10–100 µm, outer diameter of 1–5 µm) composed of nanoparticles (average size of 15–20 nm) with a surface area of 80–90 m2 g–1 and pore volume of 0.294–0.372 cm3 g–1. The effect of Hf ion incorporation on the electrochemical behaviour of anatase TiO2 in the Li-ion battery anode was investigated by galvanostatic charge/discharge and electrochemical impedance spectroscopy. It was established that Ti0.95Hf0.05O2 shows significantly higher reversibility (154.2 mAh g–1) after 35-fold cycling at a C/10 rate in comparison with undoped titania (55.9 mAh g–1). The better performance offered by Hf4+ substitution of the Ti4+ into anatase TiO2 mainly results from a more open crystal structure, which has been achieved via the difference in ionic radius values of Ti4+ (0.604 Å) and Hf4+ (0.71 Å). The obtained results are in good accord with those for anatase TiO2 doped with Zr4+ (0.72 Å), published earlier. Furthermore, improved electrical conductivity of Hf-doped anatase TiO2 materials owing to charge redistribution in the lattice and enhanced interfacial lithium storage owing to increased surface area directly depending on the Hf/Ti atomic ratio have a beneficial effect on electrochemical properties.

The effect of Hf ion incorporation on the electrochemical behaviour of anatase TiO 2 in the Li-ion battery anode was investigated by galvanostatic charge/discharge and electrochemical impedance spectroscopy. It was established that Ti 0.95 Hf 0.05 O 2 shows significantly higher reversibility (154.2 mAh g -1 ) after 35-fold cycling at a C/10 rate in comparison with undoped titania (55.9 mAh g -1 ). The better performance offered by Hf 4+ substitution of the Ti 4+ into anatase TiO 2 mainly results from a more open crystal structure, which has been achieved via the difference in ionic radius  . . in the ionic radius values of Ti 4+ and M n+ leads to changing crystal lattice parameters of TiO 2 after doping that may result in the facilitation or slowing of Li + ions' diffusion kinetics. Thus, in order to design an efficient anatase TiO 2 anode for high power density, high-safety LIBs, the balancing between the ionic radius and the oxidation number of the dopant is a key factor. Hence, a clear understanding of the doping strategy from the point of view of the M n+ ionic radius is strongly required.
Here, the doping with Hf 4+ of nanostructured anatase TiO 2 tubes by an inexpensive template sol-gel method is reported. The relationship between the electrochemical behaviour of Ti 1-x Hf x O 2 (x = 0.01; 0.03; 0.05) in the LIB anode and the ionic radius of a substitutional agent is investigated by charge/discharge tests and electrochemical impedance spectroscopy. Based on the results in this work coupled with the data published in our previous report [28], the importance of the dopant ionic radius is discussed in detail and a good grasp of the principles of TiO 2 doping is achieved. 2. Experimental section 2. 1

. Synthesis procedure
Hf-doped anatase TiO 2 in the form of nanoparticle-structured tubes was synthesized by a template solgel procedure, which had been developed by us earlier [29]. Analytical grade titanium tetrachloride (Component-Reaktiv, Russia) and hafnium oxychloride hydrate (Sigma-Aldrich, USA) were used as precursors without any purification. As a template, Busofit-T055 carbon fibre (Khimvolokno, Belarus) was applied. As the Busofit-T055 fibre contains approximately 0.01% silicon as an impurity, the preliminary autoclave treatment with NH 4 HF 2 at 130°C was carried out. As a result, silicon concentration decreases by 30 times. In a typical synthesis process, 0. 5

Characterization
The particles' surface morphology was observed by scanning electron microscopy (SEM) on a S5500 microscope (Hitachi, Japan) The samples for SEM were prepared by spreading the undoped or Hf-doped anatase TiO 2 on sticky conductive adhesive tape. Additionally, the microstructure was investigated by transmission electron microscopy (TEM) on a Titan 80-300 (FEI, USA) equipped with a spherical aberration (Cs) corrector of an electron probe. The TEM was operated under an acceleration voltage of 300 kV in a bright-field imaging and high-angular dark field scanning (HAADF STEM) mode. For (S)TEM observations, the specimens were dispersed during 5 min in the ultrasonic bath with ultrapure water, which was obtained using a Milli-Qwater purification system (Millipore, USA). Then, a drop of suspension was applied to a copper grid with a lacy carbon support film. TEM samples were treated for 20 s in a Model 1020 plasma cleaner (Fischione, USA) using the Ar/O 2 gas mixture to reduce carbohydrate contamination. The GATAN DIGITAL micrograph software was used for image processing and analysis. STEM image simulation was performed with the JEMS electron microscopy software. Energy-dispersive X-ray (EDX) microanalysis was performed on the Versa 3D SEM (FEI, USA) equipped with the high-count rate silicon drift detector Octane-plus (EDAX, USA). The accelerating voltage for EDX mapping was set to 15 keV and the beam current was 1.7 nA. The specific surface area was determined using the ASAP 2020 V3.04 H (Micrometrics, USA) spectrometer from isotherms of lowtemperature adsorption of nitrogen by the Brunauer-Emmett-Teller (BET) method. The pore diameter and total volume of pores were evaluated using the original density functional theory. The surface chemistry was revealed by X-ray photoelectron spectroscopy (XPS) on a Phoibos 150 hemispherical electrostatic energy analyser (SPECS, Germany). Mg K α -radiation was used as the primary excitation source. The measurements were refined for possible charging effects by assigning a value of 285.0 eV to the C 1 s reference line resulting from the thin layer of residual hydrocarbon. To investigate the crystal structure of materials, X-ray diffraction (XRD) and Raman spectroscopy were used. The D8-Advance diffractometer (Bruker, Germany) with Cu K α -radiation was applied for XRD measurements. Identification of the XRD data was performed using the EVA program with the PDF-2 (2006) powder database. Raman studies were conducted on the RFS-100/S spectrometer (Bruker, Germany) equipped with a Ge detector. As the excitation source an Nd:YAG laser with a wavelength of 1064 nm was applied. Thermogravimetry (TG) was carried out on the DTG-60H derivatograph (Shimadzu, Japan) at a heating rate of 5°C min -1 under an air atmosphere from room temperature to 1000°C.

Electrochemical tests
The working electrode was composed of anatase Ti 1-x Hf x O 2 (x = 0; 0.01; 0.03; 0.05) as an active material, Super P carbon black (Alfa Aesar, USA) as a conductive additive and polyvinylidene fluoride (MTI, USA) as a binder at a weight ratio of 80 : 10 : 10. The mixture was homogenized in the N-methylpyrrolidone solvent (Ekos-I, Russia) using the C-MAG HS 7 magnetic stirrer (IKA, China) at a rate of 300 r.p.m. for 15 h to prepare a homogeneous viscous slurry. Then, the electrode slurry was spread onto a copper current collector sheet (thickness is 11 µm) by the doctor blade method using the AFA-I instrument (MTI, USA). The electrode sheet was dried at 70°C for 7 h in the DZF-6020-110P oven (MTI, USA). The T06 tool (MTI, USA) was used for the cutting of a round electrode disc (diameter is 1.5 cm) from the sheet. Finally, the working electrode was compacted at 1000 kg cm -2 on a C3851 press (Carver, USA) and dried under vacuum at 110°C overnight. The mass loading of active material was approximately 2 mg cm -2 .
The half-cell was assembled in a 890-NB glove box (Plas-Labs, USA) under a dry (H 2 O < 1 ppm) Arfilled (purity is 99.999%) atmosphere. A two-electrode ECC-STD cell (Bio-Logic, USA) was applied to test the electrochemical performance. Lithium metal (Lithium-Element, Russia) was used as a counter and reference electrode. A 1 M solution of LiClO 4 in the mix of propylene carbonate and dimethoxyethane at a volume ratio of 5 : 1 (Ekotech, Russia) was applied as the electrolyte. To prevent short circuit, a Celgard 2400 polypropylene separator (Celgard, USA) was used.
Galvanostatic cycling tests at the rates of C/10 and 1C (C is equal to 335 mA g -1 ) were performed on a 1470E potentiostat/galvanostat (Solartron, UK) between the potentials of 1.0 and 3.0 V. In the experiments, because of applying half-cells, the discharge implies a lithiation process, while the charge implies de-lithiation. Electrochemical impedance spectroscopy (EIS) data were collected on a 1455 (Solartron, UK) frequency response analyser at room temperature for the fresh cells at opencircuit potential with an AC amplitude of 5 mV over a frequency range from 1 MHz to 100 mHz. The measurements were carried out on at least six half-cells for each test. 3. Results and discussion 3 Additionally, TEM analysis demonstrates the structure of Ti 0.95 Hf 0.05 O 2 presented by nanoparticles with inter-particle pores. The N 2 adsorption isotherm and the corresponding evaluation of pore diameter and total volume of pores further confirm the porous microstructure of the Hf-doped TiO 2 . The average pore diameter for undoped titania is 1.48 nm and the total pore volume is 0.294 cm 3 g -1 . Ti 0.95 Hf 0.05 O 2 possesses a pore diameter of 3.17 nm and total pore volume of 0.372 cm 3 g -1 . Furthermore, the Hf 4+ doping of anatase titania increases the BET surface area from 80 m 2 g -1 (undoped TiO 2 ) to 90 m 2 g -1 (Ti 0.95 Hf 0.05 O 2 ), which seems due to the slightly reduced particle size. According to the literature [30], porosity and surface area have a beneficial effect on the electrochemical activity of anatase TiO 2    As is well known, the imaging in the HAADF STEM mode is atomic number-sensitive, so the contrast is proportional to Z 2 . Hence, owing to a large difference in atomic numbers (Z Hf = 72 versus Z Ti = 22), Hf atoms appear much brighter than the Ti species. Indeed, STEM imaging of Ti 0.95 Hf 0.05 O 2 in the Z-contrast mode (figure 2a) reveals that, in some cases, Hf atoms occupy the positions of titanium in the anatase TiO 2 crystal lattice. One of those Hf atoms is marked with a yellow circle in figure 2a, inset with higher magnification. Simulations of HAADF STEM images were performed to compare the difference in contrast presented at experimental images. The parameters of the microscope were set close to experimental ones. Two single Hf atoms were placed in the titanium atoms' positions. Figure 2b and c, respectively, represents an experimental and simulated structure of anatase TiO 2 in the (100) orientation, where the unit cell c direction aligned vertically. For qualitative interpretation, the intensity line profiles are shown in both images. It can be concluded that simulated data are in a good agreement with experimental ones and it is obvious that Hf atoms are incorporated into the lattice in the position of titanium atoms. In this study, we did not aim to visualize the oxygen atoms in anatase TiO 2 structure, as it was previously reported [31]. Moreover, in the case of the nanoparticles, it is rather difficult to obtain exactly the predetermined zone axis, and also the weak scattering intensity of oxygen atoms makes them invisible in the presented micrographs.
According to the elemental mapping, Ti, Hf and O elements (electronic supplementary material, figure S1) are uniformly distributed in Ti 0.95 Hf 0.05 O 2 . Thus, Hf is incorporated homogeneously into the anatase TiO 2 crystal lattice. Moreover, EDX measurements give the Hf/Ti atomic ratio to be equal to    3) shows that Ti, O, Hf and C elements are clearly detected in the material, which is in accordance with EDX data. By contrast, Si was not found, which seems to be owing to a lower penetration depth (30 Å) of the XPS method in comparison with EDX (approx. 1 µm). The high-resolution spectrum of Ti 2p (electronic supplementary material, figure S2a) shows two peaks of Ti 2p3/2 at 459.1 eV and Ti 2p1/2 at 464.8 eV, indicating that titanium is present as Ti 4+ [32]. Note that, for Hf-doped TiO 2 , there are no signals of Ti 3+ (approx. 455.0 eV 27 ) in the Ti 2p spectrum results. This seems to indicate that the Ti oxidation state has not been changed owing to Hf doping. The O 1 s peak (electronic supplementary material, figure S2b) exhibits two main contributions. In particular, the binding energy of 530.6 eV belongs to oxygen atoms of the TiO 2 lattice [33], whereas the band at 532.5 eV corresponds to chemisorbed water, and C−O and O=C−O bonds [34]. From the XPS results (electronic supplementary material, table S1), note that the O/Ti atomic ratio is close to 2 (46.0/20.5). The Hf 4f7/2 and Hf 4f5/2 peaks located at 16.8 Ti 0 -1 ) of B 1 g(1) . The high-intensity E g(1) peak undergoes the greatest shift (from 147.5 to 143.4 cm -1 ) because of its sensitivity to changes in unit-cell parameters  [40]), the SiO 2 is amorphous. At the same time, it should be noted that all samples displayed the typical low-intensity D (1305 cm -1 ) and G (1590 cm -1 ) peaks (figure 5, inset) associated with the annealed carbon template in the amorphous and graphitized states [41].
To evaluate the concentration of carbon originated from the pyrolysis of Busofit-T055 fibre, thermogravimetric analysis (TGA) was performed. As shown in figure 6, the TGA curves of Ti 1-x Hf x O 2 (x = 0; 0.01; 0.03; 0.05) samples depict three steps of weight loss. The first region below 200°C is associated with the release of H 2 O adsorbed on the TiO 2 surface [42]. The second interval from 200 to 650°C is related to the removal of residual carbon because the Busofit-T055 carbon fibre is fully decomposed at approximately 650°C (figure 6, inset). The third weight loss step between 650 and 850°C can be attributed to the dehydration of surface Ti(OH) 2 [43]. Thus, the overall weight loss for all samples varies between 2.5% and 5.5%.  same time, Ti 0.95 Zr 0.03 O 2 , which was the best sample, exhibited a reversible capacity of only 135 mAh g -1 after 35 cycles at a C/10 rate. Such worse performance can be explained by the slightly larger ionic radius of Zr 4+ (0.72 Çž) when compared with that of Hf 4+ , which may cause the undesirable excessive lattice strain in anatase TiO 2 . This suggestion is also confirmed by our unsuccessful efforts to increase the Zr to Ti ratio to more than 0.03. Overall, the capacity of nanostructured Hf-doped anatase TiO 2 tubes is higher than that of Nbdoped TiO 2 nanofibres (128 and 92 mAh g -1 for the 20th cycle at rates of C/20 and C/5, respectively), as demonstrated by Fehse et al. [44]. On the other hand, Wang et al. [32] reported an enhanced cycling performance (160 mAh g -1 after the 100 cycles at a C/6 rate) for mesoporous Nb-doped anatase titania with a specific surface of 128 m 2 g -1 . Hence, the relationship between the synthesis technique as well as morphology and electrochemical behaviour of the material is very strong. Note that, in the present work, the direct comparison of nanostructured undoped and Hf-doped anatase titania tubes reveals a superior cycling performance of the doped samples.

Electrochemical performance of Ti 1-x Hf x O 2
To understand the reasons of higher performance behaviour of Hf-doped anatase TiO 2 electrodes, the EIS method was applied. As shown in figure 8, the Nyquist plot consists of the high-frequency semicircle and the low-frequency arc. The EIS spectra have been fitted (table 2) using an equivalent circuit (figure 8, inset) composed of internal resistance R s , charge transfer resistance R ct at the double layer CPE dl (and, possibly, Li + migration through the SEI) and Warburg impedance Z w associated with diffusion of Li + ions into the solid phase. The collected data demonstrate that the incorporation of Hf 4+ ions into the where R is the gas constant (J mol −1 K −1 ), T is the temperature (K), S is the contact area between the electrode and electrolyte (cm 2 ), n is the charge transfer number, F is the Faraday constant (C mol −1 ) and C Li is the concentration of Li + in TiO 2 (mol cm -3 ). The Warburg factor was found (table 2) from the slope of the fitting line in the plot of the real axis impedance values Z versus the reciprocal square root of the angular frequencies ω -1/2 (electronic supplementary material, figure S3) according to the following relationship: Figure 9 reveals the reversibility of Li + ion intercalation/de-intercalation into/from the Ti 0.95 Hf 0.05 O 2 electrode after its cycling at different C-rates. Ti 0.95 Hf 0.05 O 2 was first cycled at C/10 and, after seven cycles, the rate was increased up to 1C. A reversible capacity of 165.0 mAh g -1 was obtained at a C/10 rate

Conclusion
In this work, a promising facile way of TiO 2 modification by Hf-doping via a template sol-gel route in order to improve its Li + storage properties is suggested. It was found that the as-synthesized As demonstrated, the Hf 4+ ions are homogeneously incorporated into the titania lattice by substitution of Ti 4+ . Owing to the incorporation of Hf ions the titania unit cell parameters are increased. When the Ti 1-x Hf x O 2 materials are used as LIB anodes, the lithiation and de-lithiation capacities as well as cycling performance of the as-synthesized TiO 2 -based materials significantly improved by Hf 4+ doping. Indeed, 35-fold charge/discharge cycling at a C/10 rate in the range 1.0-3.0 V shows that the reversible capacity for Hf-doped TiO 2 is significantly higher (e.g. 154.2 mAh g -1 for Ti 0.95 Hf 0.05 O 2 ) than that for the undoped titania (55.9 mAh g -1 ). The Hf 4+ substitution of the Ti 4+ in TiO 2 structure leads to the charge transfer resistance R ct decreasing (62.1 versus 280.1 Ω for Ti 0.95 Hf 0.05 O 2 and undoped TiO 2 respectively), which seems to be because of enhancement of conductivity. Moreover, the diffusion coefficient D Li of Li + ions is increased by almost three orders of magnitude from 7.33 × 10 -17 cm 2 s -1 (TiO 2 ) up to 6.91 × 10 -14 cm 2 s -1 (Ti 0.95 Hf 0.05 O 2 ). Overall, the better reversibility of the electrochemical process occurring in Hf-doped anatase TiO 2 electrodes is associated with: (i) changing of the unit cell parameters, which facilitates Li + ion diffusion; (ii) charge redistribution in the lattice, which improves conductivity of TiO 2 ; and (iii) increased surface area, which enhances interfacial lithium storage. Comparison of the obtained results with those for Zr-doped anatase TiO 2 published earlier demonstrates that ionic radius of dopant M n+ partially substituted with Ti 4+ in the anatase TiO 2 lattice is a key factor for the design of advanced anodes for high-safety LIBs.