Article pubs.acs.org/cm
Cite This: Chem. Mater. 2017, 29, 10007−10018
Chlorine Doping of Amorphous TiO2 for Increased Capacity and Faster Li+‑Ion Storage Sébastien Moitzheim,†,‡ Joan Elisabeth Balder,§ Paul Poodt,§ Sandeep Unnikrishnan,§ Stefan De Gendt,†,∥ and Philippe M. Vereecken†,‡ †
imec, 3001 Leuven, Belgium KU Leuven, Department of Microbial and Molecular Systems, Centre for Surface Chemistry and Catalysis, 3000 Leuven, Belgium § TNO-Holst Centre, 5656 AE Eindhoven, The Netherlands ∥ KU Leuven, Department of Chemistry, Molecular Design and Synthesis, 3000 Leuven, Belgium
Chem. Mater. 2017.29:10007-10018. Downloaded from pubs.acs.org by UNIV OF SOUTH DAKOTA on 09/05/18. For personal use only.
‡
S Supporting Information *
ABSTRACT: Titania (TiO2) offers a high theoretical capacity of 336 mAh g−1 with the insertion of one Li per Ti unit. Unfortunately, the poor ionic and electronic conductivity of bulk TiO2 electrodes limits its practical implementation. Nanosizing titania below ∼20 nm has shown to increase the rate performance and accessible capacity but still not more than 75% of the theoretical capacity at 1 C. In this work, we discovered that chlorine doping of amorphous TiO2 (TiO2−xCl2x) can achieve a high capacity without the need for nanosizing. By in situ doping during atomic layer deposition, an unprecedented 90% of the theoretical capacity was achieved at 1 C for 100 nm thick films. Even at a charging rate of 20 C, 40% of the maximum capacity was accessible for the film with highest Cl-content (x = 0.088). The capacity was found linearly dependent on the chloride content for a Cl/Ti atomic ratio from 0.06 to 0.09. The enhanced insertion kinetics are ascribed to enhanced electronic conductivity and facilitated Li+-ion diffusion as a result of Cl-doping. Furthermore, the potential of TiO2−xCl2x films as high rate anode were demonstrated on micropillar electrodes in a half-cell configuration using a liquid electrolyte solution, showing 10 times higher capacity at 10 C compared to the literature. carbon composites8,12,13 and amorphous TiO2 (am-TiO2) nanotubes14,15 have shown the most promising fast charging results, with capacities reaching 150 mAh g−1 (45%) at 50 C for the amorphous TiO2 nanotubes.14,15 Note, however, that the high rate performance of amorphous TiO2 so far only has been reported for nanosized dimensions. Although nanosizing has proven a powerful tool to improve battery electrodes, such systems are still plagued by increased solid-electrolyte interphase formation, reduced packing density and high cost.16 Additionally, electrodes with submicron (≥100 nm) thicknesses are envisioned for three-dimensional (3D) thin-film solid-state batteries.17 In 3D thin-film batteries, a stack of cathode, electrolyte, and anode thin-films are conformal coated over a 3D microstructured current collector substrate.18 To deposit the submicron thick am-TiO2 films, we chose spatial atomic layer deposition (S-ALD). This technique can produce high quality, smooth, and conformal films as for conventional temporal ALD, but at a fraction of the time,19 and can potentially be scaled in a roll-to-roll manufacturing process.20 For example, a
1. INTRODUCTION The fast development of portable electronic devices, such as smartphones, autonomous sensors, and wearable electronics, leads to increasing demands for energy storage. To meet such demands, new Li+-ion battery materials are being developed that offer higher storage capacities and charging rates. An electrode material that has received considerable attention as potentially cheap, environmentally friendly, and stable Li+-ion insertion material is TiO2. Theoretically, a high capacity of 336 mAh g−1 can be achieved by insertion of 1 Li per TiO2 unit formula with the reduction of Ti(IV) to Ti(III). Unfortunately, its low electronic and low Li+-ion conductivity result in a poor rate-performance and, in turn, achievable capacity. Typically, for “bulk” anatase, 0.6 Li per TiO2 can only be realized at very low charging rates below 0.01 C. To improve the rateperformance and capacity of TiO2, many strategies have been explored. These include alternative TiO2 crystal structures,1−3 nanostructuring,4−6 carbon based composites,7,8 and doping. 9−11 Even for the best performing nanosized TiO 2 electrodes, the accessible capacity at 1 C (i.e., 336 mA g−1 or current to reach LiTiO2 in 1 h) is still, in the best case, limited to about 250 mAh g−1 (or Li0.74TiO2). An overview of reported capacities for state-of-the-art TiO2 electrodes is provided in Table S1. Among these reports, nanosized amorphous TiO2/ © 2017 American Chemical Society
Received: August 17, 2017 Revised: November 12, 2017 Published: November 13, 2017 10007
DOI: 10.1021/acs.chemmater.7b03478 Chem. Mater. 2017, 29, 10007−10018
Article
Chemistry of Materials
mounted.21 Precursor inlets are surrounded by exhaust zones and are incorporated in a 150 mm diameter round reactor head situated above the sample. The inlets are surrounded by gas bearing planes, which separate the different reaction zones and prevent precursor intermixing. Gas bearing is formed by flowing pressurized N2 through holes located on the gas bearing surface. The sample table can be rotated at different rotation frequencies. The entire construction is mounted in a convection oven which controls the deposition temperature. More details on the reactor used can be found in ref 21. Depositions were done simultaneously on four 2 × 2 cm2 samples which were mounted on a holder and loaded into the S-ALD reactor. For each deposition run, two planar and two 3D substrates were loaded. The deposition precursors were TiCl4 and H2O, which were evaporated from a bubbler using 50 and 500 sccm flow through the bubblers, respectively. Both flows were further diluted by N2 resulting in total volume flows (N2 + precursor) of 1 slm (standard liter per minute). The temperature of the precursor bottles was controlled outside the reactor and kept at room temperature for TiCl4 and at 50 °C for H2O. Depositions with 20, 30, and 40 rpm substrate rotation frequencies were performed, which correspond to 140, 90, and 70 ms of gas exposure time at the center of the sample, respectively. As a chlorine-free reference, 100 nm am-TiO2 samples were deposited by SALD at 100 °C and 280 ms exposure time, using titanium tetraisopropoxide (TTIP) and H2O as precursor. The longer exposure time for TTIP (vs TiCl4) is needed to ensure self-limiting growth. 2.3. Structural Characterization. The thickness and surface morphology of TiO2−xCl2x films deposited on planar TiN/Si and micropillar substrates were examined with a NOVA 200 (FEI) scanning electron microscope (SEM). Grazing incidence XRD (GIXRD) was performed with an X-Pert PRO MRD (Panalytical), equipped with a Cu Kα X-ray source. Measurements were done at an incidence angle of 1°, and the signal was continuously recorded by scanning in the 2θ direction at a rate of 0.025°/s. The chemical state and content of chlorine in the TiO2−xCl2x layers were determined using X-ray photoelectron spectroscopy (XPS) and Rutherford backscattering spectroscopy (RBS). XPS was carried out with the Quantera tool from ULVAC-PHI (Q1). Prior to XPS measurements, surface contaminations of the samples were removed by a 2 min sputter clean at 1 keV with Ar+ ions. For RBS measurements, a He+ beam is accelerated to an energy of 1.52 MeV and scattered off the film. The backscattered ions are detected by a time-of-flight energy telescope, which give information on the elemental composition.29 More information on the RBS setup used in this study can be found in ref 29. 2.4. Electrochemical Characterization. A custom-made threeelectrode polytetrafluoroethylene (PTFE) cell (see Figure S2) was used, which is clamped onto the substrate using a Kalrez O-ring, with a surface area of 1.1 cm2 and 1.79 cm2 for the planar and 3D substrates, respectively. Both planar and 3D sample used the same electrochemical cell, but with a larger O-ring for the 3D sample to accommodate the 1 × 1 cm2 pillar array. The cell contains two compartments, one compartment comprising a Li metal foil as a counter and the other comprising a Li metal foil as a reference electrode. The compartment with the Li reference electrode was connected to the main compartment through a Luggin capillary close to the surface of the working electrode (at ∼4 mm), and the cell was filled with a liquid electrolyte solution. All experiments were performed at room temperature (21 °C) using a LiClO4 in propylene carbonate (PC) electrolyte solution with our three-electrode cell. For ease of preparation, an ampule containing LiClO4 (100 g, battery grade, dry, 99.99%, Sigma-Aldrich) was dissolved in propylene carbonate (100 mL, 99.7%, Sigma-Aldrich), which leads to a 0.94 M solution. Measurements were done in an Ar filled glovebox with O2 and H2O kept below 1 ppm. Electrical contact was made to the samples by scratching the back of the sample and applying a gallium indium eutectic (Alfa Aesar) and contacting with Cu foil. The electrochemical cell was controlled through a PGSTAT101 Autolab (Metrohm) potentiostat/galvanostat, using the Nova 1.10 software. Five cyclic voltammogram (CV) cycles were recorded at 10 mV s−1 in the range of 0.1−3.2 V, after which galvanostatic lithiation/delithiation
100 nm Al2O3 passivation layer for photovoltaic cell can be deposited in 90 s, while this would last 2.5 h with conventional ALD.21 In our case, we could deposit 100 nm am-TiO2 in 30 min to 1 h, while we needed 19 h for a similar coating using conventional ALD. As the precursor for S-ALD am-TiO2 deposition, we initially used titanium isopropoxide (TTIP) and H2O. However, submicron thick am-TiO2 deposited using these precursors did not deliver satisfactory capacity and fast charging rate capabilities. Therefore, we investigated if doping of am-TiO2 could enhance the electrode performance. For this, TiCl4 and H2O were used as S-ALD precursors instead, which is known from ALD literature to result in chlorine incorporation when a sufficiently low deposition (≤130 °C) temperature is chosen.22,23 By doing this, we discovered that chlorine doping of amTiO2 can improve the capacity multifold, especially at high charging rates. Previously, reports on doping have all focused on crystalline TiO2 (e.g., with doping elements such as Nb, Ni, Fe, V, Mn, N, S, and F) which improved the capacity and fast charging capability, attributed mainly to increased electronic conductivity.11,24−27 To date, there are no reports on Cl-doped TiO2 (amorphous or crystalline) and in general no reports on doped am-TiO2. In this work, we systematically studied the influence of Cl content on the lithium ion insertion and extraction capacities of am-TiO2 electrodes and compared the results to chlorine free and other reported TiO2 based electrodes. Layers 100 nm in thickness were characterized by grazing incidence XRD (GI-XRD), Rutherford backscatter spectroscopy (RBS), X-ray photoelectron spectroscopy (XPS), and electrochemical testing (cyclic voltammograms and galvanostatic charge/discharge). Furthermore, am-TiO2 layers were deposited by S-ALD on high-aspect ratio microstructures (TiN-coated Si micropillars) to demonstrate Cl-doped amTiO2 as a suitable electrode and S-ALD as a potentially scalable technique for three-dimensional (3D) thin-film batteries. Finally, our best performing 3D electrode is shown to outperform all other TiO2-based 3D thin-film electrodes reported, which shows the advantage of chlorine doped amTiO2 and its potential as the anode for all-solid-state 3D thinfilm batteries.
2. EXPERIMENTAL SECTION 2.1. Planar and Micropillar Substrates. Planar Substrates. A 200 mm n-type phosphorus doped crystalline Si wafer was coated with 60 nm TiN by physical vapor deposition and diced in 2 × 2 cm2 pieces. The samples were used as-is for S-ALD am-TiO2 deposition and subsequent electrochemical characterization. TiN was chosen as the current collector and diffusion barrier for lithium into the Si.28 Growth per cycle determination of the S-ALD am-TiO2 process was performed with am-TiO2 layers deposited on bare 152 mm Si wafers. Silicon Micropillar Substrate. A 300 mm n-type phosphorus doped Si wafer was used for the Si micropillar fabrication. Standard photolithographic patterning, combined with deep reactive ion etching were employed to fabricate the pillar structures. Si micropillar arrays were defined on the 300 mm wafer in 1 × 1 cm2 squares, with 1 cm spacing in-between the arrays. Pillars were ordered in a square lattice, with a diameter and interpillar spacing of 2 μm and a (nominal) height of 50 μm (see Figure S1). A 23 nm TiN current collector was deposited on the pillars by conventional ALD using a plasma based process at 370 °C. Planar and micropillar substrates used in SEM and electrochemical characterization were diced in 2 × 2 cm2 squares (with the micropillar array in the center) on which S-ALD of am-TiO2 was performed. 2.2. Spatial Atomic Layer Deposition of TiO2. A rotary type reactor was used which allows up to 152 mm round substrates to be 10008
DOI: 10.1021/acs.chemmater.7b03478 Chem. Mater. 2017, 29, 10007−10018
Article
Chemistry of Materials
Table 1. Ti Content, Cl/Ti Atomic Ratio, Film Density, and Theoretical Capacity As Measured by RBS for 100 nm Amorphous TiO2 Films Deposited by S-ALD at Different Deposition Conditions (Temperature/Exposure Time) and Gas Precursors Ti content (1016 atomic cm−2)
Cl/Ti atomic ratio
densitya (g cm−3)
theoretical capacity (mAh cm−3)b
TiCl4 + H2O
100 115 115 115 130
°C/70 ms °C/70 ms °C/90 ms °C/140 ms °C/90 ms
21 21 21 21 27
0.088 0.074 0.060 0.059 0.064
2.93 2.91 2.90 2.90 3.18
930 930 930 930 1200
TTIP + H2O
100 °C/280 ms
21
0
2.84
930
precursors
a
temperature/exposure time
On the basis of the TiO2−xCl2x stoichiometry. bCalculated from the Ti content for the reduction of Ti(IV) to Ti(III).
experiments were carried out with cutoff voltages of 3.0 and 0.1 V. Between CVs and lithiation/delithiation experiments, the electrode was relaxed at 3.0 V until a 1/50 C cutoff current was reached. Specifically for the planar samples, a current of 5, 9, 23, 50, 90, 230, 500, 2, and 5 μA cm−2 was consecutively applied. These current densities correspond to a C-rate of approximately 0.5, 1, 2.5, 5, 10, 25, 50, 0.25, and 1 C, respectively, based on the theoretical capacity of our am-TiO2 films (i.e., 930 mAh cm−3 or 9.3 μAh cm−2, see Table 1). For the 3D samples, a current density of 100, 200, 500, 1000, 2000, 5000, 50, and 100 μA cm−2 were applied in sequence, which corresponds to a C-rate of 0.5, 1, 2.5, 5, 10, 25, 0.25, and 0.5 C, respectively. Potentialdependent electrochemical impedance spectroscopy (PEIS) was performed at different potentials between 0.8 and 2.8 V. Prior to the PEIS experiments, samples were precycled 10 times using CV scans between 0.1 and 3.2 V at 10 mV s−1. After application of the desired potential, the cell was relaxed until a current of 100 nA cm−2 was reached before the start of each impedance measurement. The impedance was recorded at frequencies ranging from 100 kHz to 5 mHz, with a perturbation amplitude of 10 mV. Long-term cycling tests were done using a 3D electrode with TiO2−xCl2x deposited at 100 °C/ 70 ms. Before long-term cycling, five CV cycles were recorded at 10 mV s−1. Subsequently, 1000 lithiation and delithiation cycles at a rate of 10 C and 5 cycles at 1 C were applied. All voltages are given versus Li+/Li.
was confirmed using grazing incidence XRD, where no diffraction features were seen, except for the underlying substrate (see Figure S4). The thickness and morphology of the deposited films on planar TiN substrates were also determined by SEM which showed, smooth, closed, and crack-free films for the layers deposited at 100 and 115 °C (see Figure S5). For the films deposited at 130 °C, however, some particulate features were visible. Because of nonuniformity of the 130 °C TiO2−xCl2x film and associated difficulties in determining the exact film density, we did not evaluate these further for electrochemical performance. 3.2. Chemical Analysis of Chlorine Doped TiO2 Films. The stoichiometry of the chlorine-doped TiO2 films was measured using Rutherford backscatter spectroscopy (RBS). Table 1 gives the results of the Ti content, Cl/Ti atomic ratio, and film density for the different deposition conditions (see Figure S6 for “raw” RBS data). An equal Ti content is measured for depositions performed at 100 and 115 °C (both for TiCl4 and TTIP), whereas the Ti content is higher for 130 °C. On the basis of the TiO2−xCl2x stoichiometry (see below), a density of about 2.9 g cm−3 is calculated for layers deposited at 100 and 115 °C, irrespective of exposure time. Note that this density is considerably lower than for example bulk anatase TiO2 (∼3.8 g cm−2).31,32 For deposition at 130 °C, a density of 3.2 g cm−3 is calculated, which is about 10% higher than the layers deposited at 100 and 115 °C and matches the observed difference in GPC. Since some overlap between the RBS Ti signal from TiN and TiO2−xCl2x occurred, a fitting error of about 5% for the Ti content was noticed. On the basis of the Ti content and by assuming the reduction of Ti(IV) to Ti(III), a theoretical capacity of 930 mAh cm−3 is calculated both for TiCl4 and TTIP-based TiO2 deposited at 100 and 115 °C. The Cl content was analyzed by RBS for the different layers deposited with the TiCl 4 and TTIP precursor. The highest Cl content (TiO1.956Cl0.088) is achieved for the lowest deposition temperature of 100 °C. Deposition at 115 and 130 °C all lead to lower Cl contents compared to 100 °C. For deposition at 115 °C, the influence of gas precursor exposure time on Cl content was investigated, which shows that a shorter exposure time leads to more incorporated Cl. Hence, both deposition temperature and exposure time can be used to control the Cl content, which will be explained using reaction mechanism discussed below. The dependence of the Cl content on the exposure time may in some cases be a disadvantage, for example, when depositing on high aspect ratio structures. In this case, the top might receive a higher flux/dose than the bottom and lead to a nonuniform Cl content as a function of penetration depth. However, from the electrochemical results of our Cl-doped am-TiO2 films deposited on micropillar structures with an aspect ratio of 25
3. RESULTS 3.1. Spatial ALD of (Chlorine Doped) TiO2 Films. Amorphous TiO2 films were deposited by Spatial ALD (SALD), which relies on the spatial rather than the conventional temporal distribution of reagent precursors.19 Depositions are performed at atmospheric pressure while preserving the selflimiting nature of conventional ALD. Substrate temperatures were fixed at 100, 115, or 130 °C, which is low enough to form amorphous TiO2 films.22,30 The growth per cycle was determined by spectroscopic ellipsometry with TiO2 layers deposited on full 152 mm (6 in.) Si wafers (see Figure S3). Because of our rotary reactor design, the effective exposure time depends on the radial distance away from the center of the reactor. Therefore, by measuring the TiO2 thickness at different locations on the wafer as a function of distance away from the center of the wafer, several exposure times can be probed in the same deposition run. For 100, 115, and 130 °C a growth per cycle (GPC) of 0.085, 0.080, and 0.075 nm/cycle was determined. At an exposure time of 70 ms and above, the GPC at each deposition temperature is found to be independent of exposure time (see Figure S3). For am-TiO2 layers deposited from the TiCl4 precursor, 1155, 1225, and 1538 rotation cycles were adopted, which corresponds to 98 nm for 100 and 115 °C and 115 nm for and 130 °C. Note that the increased GPC for depositions performed at lower temperatures suggests that a lower temperature leads to less dense films. The amorphous nature of the Cl-doped TiO2 films 10009
DOI: 10.1021/acs.chemmater.7b03478 Chem. Mater. 2017, 29, 10007−10018
Article
Chemistry of Materials
not allow us to vary the exposure time for the TiCl4 and H2O precursor separately, as they are both determined by the substrate rotation frequency. On the other hand, the observed Cl/Ti ratio might be the result of a chemical equilibrium forming at the specific deposition temperature. In this case, increasing the H2O dose would not lead to changes in the Cl/Ti ratio. Increasing the H2O could be accomplished by increasing the vapor pressure of the H2O precursor, but this has at present not been investigated. More complex mechanisms, like nongrowth ligand exchange reactions, can also take place, such as
(see section 3.4), there is no significant difference in Li-ion storage performance between a planar and a film deposited on a 3D substrate, which shows that, in our case, the compositional variation across the length of the micropillar structure is insignificant. For the TTIP-based amorphous TiO2 layers, no Cl was measured by RBS, as there is no chlorine present in the precursors. The Cl/Ti ratios measured here are in line with previous reports of ALD TiO2: one report gave a Cl/Ti ratio of 0.055 and 0.047 for films deposited at 100 and 120 °C, respectively,33 and another a Cl/Ti content of 0.065 for deposition at 90 °C.23 Note that, although the incorporation of Cl during ALD of TiO2 with the TiCl4/H2O process has been observed before, its beneficial effect for amorphous TiO2 as Li+ion insertion electrode has never been investigated. The specific binding of Ti and Cl within am-TiO2 for films grown at 100 and 130 °C was investigated by XPS (Figure 1).
M−OH(s) + TiCl4(g) → M−Cl(s) + Ti(OH)Cl3(g) (3)
The species formed in this reaction are mobile species that can also migrate on the surface of the material.36 According to Aarik et al., these species are only observed at deposition temperatures of 200 °C and higher.37 A possible variation could be the competition reaction of gaseous HCl, formed during either H2O or TiCl4 pulse, with surface hydroxyl groups formed: M−OH(s) + HCl(g) → M−Cl(s) + H 2O(g)
resulting in the formation of inactive sites for TiCl4 pulse and incorporation of chlorine in the film. A study by Leem et al. shows that addition of a HCl pulse after the TiCl4 pulse to the ALD cycle decreases the thickness of the layer, thus decreasing the number of active sites.38 From RBS it is apparent that at increased pulse and purging times (these are related in the current setup) and increased temperature, the amount of chlorine incorporated decreases. Since the Cl content increases as the exposure time decreases, the reaction of with HCl (eq 4) is unlikely. From literature it is known that the kinetically slowest half-cycle is the removal of the Cl-ligand during the H2O pulse.39,40 Therefore, it is most likely that incorporation of Cl is a result of an incomplete ligand exchange reaction (eq 2), which explains the temperature and exposure time dependence. Assuming that Cl is present as a result of an incomplete exchange reaction mechanism, (eqs 1−4), it is expected that when Cl (with oxidation state 1−) is incorporated in the structure, Ti will remain in the 4+ oxidation state. For electroneutrality to hold, two Cl− atoms will substitute one oxygen atom, giving rise to the stoichiometry of TiO2−xCl2x. Interestingly, this implies that Cl doping will not come at the cost of maximum lithium insertion capacity, since the electrochemically active Ti(IV) states are retained. A previous study on fluorine-doped anatase TiO2 reported the stoichiometry of Ti1−x−y□(x+y)O2−4(x+y)F4x(OH)4y.9 In this case, both F− and OH− were shown to result in the creation of Ti4+ vacancies (i.e., “□”) and which in turn significantly increased the rate-performance for Li+-ion insertion. In the case of F−-doping, the overall electroneutrality is satisfied without formation of Ti3+ states. The main improvement in performance was attributed to the formation of additional diffusion channels for Li+-ion transport and modification of the insertion mechanism from a two-phase transition to a solid-solution reaction.9 In their case, doping of anatase with F− (and OH−) was shown to form well-defined vacancy sites in the crystal structure, for which the use of a vacancy symbol (□) in the chemical formula notation is well founded. In our case, however, since no long-range order exists in am-TiO2, the use
Figure 1. XPS results of the Cl- (a) and Ti- (b) 2p energy regions for an am-TiO2 films deposited at 100 and 130 °C.
The Cl 2p energy region (Figure 1a) shows a peak around 199 eV, corresponding to the Ti−Cl bonding energy.34 The shoulder apparent in the spectrum is due to the orbital splitting in 3p3/2 and 3p1/2. For similar layer thicknesses, the amount of chlorine is higher at lower temperatures, in accordance with RBS measurements. This implies that more chlorine is bound to titanium at lower temperatures. The Ti 2p energy region (Figure 1b) shows peaks related to Ti 2p1/2 at 465 eV and Ti 2p3/2 at 459 eV and suggests that the Ti is at the oxidation state of 4+ within am-TiO2.35 If Cl− would substitute for O2− in the TiO2 structure, formation of Ti3+ would be expected, but this is not confirmed by our XPS results. On the other hand, incorporation of Cl− in the structure while retaining the Ti4+ is possible according to the mechanism discussed below. The Ti−Cl bond existence in the films (Figure 1a) can be explained by several mechanisms. Ideal growth of the film would follow the following reaction paths: n(M−OH)(s) + TiCl4(g) → (M−O−)n TiCl4 ‐ n(s) + nHCl(g)
(1)
(M−O−)n TiCl4 ‐ n(s) + (4 − n)H 2O(g) → (M−O−)n Ti(OH)4 − n (s) + (4 − n)HCl(g)
(4)
(2)
A first possible explanation of the presence of persistent Cl could be an incomplete ligand exchange reaction of −Cl with −OH (eq 2). This can happen when the H2O dose (i.e., combination of H2O partial pressure and exposure time) is insufficient to fully remove the surface bound Cl-groups. In this case, an exposure-dependent Cl-content is expected, as was observed in our case. Unfortunately, our reactor design does 10010
DOI: 10.1021/acs.chemmater.7b03478 Chem. Mater. 2017, 29, 10007−10018
Article
Chemistry of Materials
fifth CV scans performed at 10 mV s−1 for the different amTiO2−xCl2x and the am-TiO2 layers are given in Figure 2a. The current response is clearly dependent on the specific deposition condition of the films and thus chloride content. For all samples, the current initially increases upon cycling and stabilizes after about 5−10 cycles (see Figure S7). This activation behavior has been observed for lithiation of other amorphous-type electrode materials47,48 and also for (undoped) amorphous TiO2.14 The largest overall current density is measured for the TiO1.956Cl0.088 film (Figure 2a). For this layer, well-defined, broad reduction and oxidation peaks are observed around 1.3 and 1.6 V vs Li+/Li, respectively, which are assigned to the Li+ion insertion and extraction into the amorphous TiO1.956Cl0.088 structure by reduction and oxidation of Ti(IV) to Ti(III). This broad shape from the voltammogram is very similar to previously reported Cl-free nanosized am-TiO28,13,49 and is related to an energetic distribution of available insertion sites within the amorphous structure.50 From Figure 2a, a lower Cl content leads to less defined reduction peaks with lower current densities. The lack of a lithiation peak suggests that the underlying kinetics are hindered by charge-transfer kinetics or other internal resistances, such as the electronic resistance of the films. As will be discussed below, the increased performance due to incorporation of Cl incorporation likely results from an enhanced electronic and ionic conductivity. Figure 2b shows the constant current lithiation and delithiation curves obtained a rate of 0.5 C (±4.5 μA cm−2) for a chlorine-free and TiO2−xCl2x films with x = 0.088 and x = 0.074. The maximum capacity is clearly dependent on the Cl content, with the highest capacity achieved for the highest Clcontent. All samples (Cl-free and -doped) show a sloped potential-capacity response. The incorporation of Cl does not give rise to additional features, except for the enhanced capacity. In line with the CV results, the sloped profile is typical for insertion and extraction in am-TiO2 and similar to previous reports on undoped am-TiO2.8,13,49 Additional lithiation and delithiation curves obtained at higher C-rates are provided in Figure S8. The rate-performance of the TiO2−xCl2x films was investigated by applying different C-rates between 0.25−50 C. The C-rates are based on the theoretical capacity determined by RBS (930 mAh cm−3, see Table 1), and a rate of x · C corresponds to a current density of about x · 9 μA cm−2. The (delithiation) capacity as a function of C-rate is shown in Figure 2c. At 0.25 C, capacities of 1030, 680, 370, and 180 mAh cm−3 are obtained for x = 0.088, 0.074, 0.060, and 0.059, respectively. The accessible capacity for x = 0.088 at higher Crates is 83% and 26% of the maximum capacity, for a rate of 1 and 50 C, respectively. Remarkably, for x = 0.088, about four times the capacity of the chlorine free am-TiO2 reference is obtained at a rate of 0.25 C. In Figure 2d, the maximum capacities obtained at 0.25 C were plotted against the Cl content. The capacity at 0.25 C was chosen as here the largest differences were observed between the different TiO2−xClx films. As a reference, the capacity of the Cl-free film is given as well. Summarizing the results above, the more chlorine in TiO2−xCl2x, the higher the Li+-ion storage capacity. Interestingly, a linear relationship is found between the Cl/Ti atomic ratio. This linear relationship is also preserved for higher C-rates (see Figure S8). The difference in capacity between x = 0.060 and 0.059 becomes smaller at higher C-rates, and at 50 C, the capacity is nearly the same. As a reference, the capacity of the Cl-free film is given as well. Summarizing the
of a vacancy notation has less meaning and was therefore omitted. Additionally, we also do not exclude the possibility of OH− inclusion in our am-TiO2 films due to the low deposition temperature used. Even though am-TiO2 lacks long-range order, short-range ordering is maintained. This is observed experimentally41 or theoretically42 by a method of pair distribution analysis. For example, ab initio modeling studies of am-TiO2 show that although the bulk structure is seemingly random, the most favorable configurations still has Ti that is 6-fold and O that is 3-fold coordinated; the same as for crystalline TiO2.42 It is likely that Cl incorporation will lead to increased local deformations, as the Ti−Cl bond length (2.21 Å) is larger than Ti−O (1.93 Å).43 In turn, this will influence different material properties such as film density and possibly Li+-ion diffusivity. Ab-initio modeling studies of nitrogen-doped am-TiO2 showed that nitrogen incorporation induces a narrowing of the electronic band gap.44,45 Similar results for Cl-doped anatase TiO2 were shown,43,46 where the band gap was decreased up to 0.95 eV for a Cl content of 4%.46 A narrowing of the bandgap will lead to enhanced electronic conductivity and will be beneficial for the lithiation and delithiation kinetics. 3.3. Electrochemical Performance of Chlorine Doped TiO2 Films. Li+-ion insertion/extraction properties of the 100 nm amorphous TiO2−xCl2x films deposited by S-ALD on planar substrates were determined by cyclic voltammetry (CV) and galvanostatic charge/discharge experiments (Figure 2). The
Figure 2. Electrochemical characterization of 100 nm films deposited by S-ALD from a TiCl4 (TiO2−xCl2x) and TTIP (am-TiO2) precursor on TiN-Si substrates with different deposition conditions. (a) Cyclic voltammetry performed at 10 mV s−1 showing the fifth cycle. (b) Galvanostatic charge/discharge experiments and resulting potential vs capacity profiles at 0.5 C (±4.5 μA cm−2). (c) Rate-performance probed between 0.25 and 50 C. (d) Delithiation capacity extracted at 0.25 C and plotted vs the Cl/Ti atomic ratio obtained from RBS. A reference line (dashed red) was taken from the Cl-free am-TiO2 data point. 10011
DOI: 10.1021/acs.chemmater.7b03478 Chem. Mater. 2017, 29, 10007−10018
Article
Chemistry of Materials
nanostructures (thin-films, nanoparticles and nanotubes), the typical distance for Li-ion diffusion into these nanostructures was taken, which is the film thickness for TiO2 films, the radius for particles, half of the wall thickness for nanotubes. This comparison assumes that the electrode performance depends only on the Li-ion and electron transport, which are strongly related to the active size of the active material. For the comparison, we use the reported average sizes of the nanostructures and gravimetric capacities (see Table S1). At a rate of 1 C, our 100 nm amorphous TiO1.956Cl0.088 film delivers 20% more capacity (301 mAh g−1) than the best performing nanosized electrode (250 mAh g−1 for am-TiO2 nanotubes with 10 nm tube wall thickness). Furthermore, at 50 C, our amorphous TiO1.956Cl0.088 achieves 92 mAh g−1, still outperforming the nanosized based TiO2 electrodes, except the am-TiO2 nanotubes (see the Supporting Information). Our hypothesis for the enhanced capacity and excellent rateperformance is based on the enhancement of both the ionic and electronic conductivity. Chlorine incorporation may improve the electronic conductivity, as it known to decrease the bandgap of crystalline TiO2.43,46 Reflectance measurements were done for our best performing TiO1.956Cl0.088 film and the am-TiO2 reference (see Figure S10) to confirm a difference in bandgap. From our reflectance measurements, a small shift in the reflectance spectra was observed between the undoped and Cl-doped film, which could be attributed to a variation in bandgap. However, more research is necessary to determine the exact influence of Cl-doping on the electronic properties. In general, the ionic conductivity of Li+ is known to be higher in amorphous TiO2 than in crystalline TiO2;53 Li-ion diffusion coefficients of 3.5 × 10−12 cm2 s−1 for amorphous54 versus ∼1 × 10−14 cm2 s−1 for anatase55 have been reported. The disordered amorphous structure of TiO2 likely allows for more facile Li+-ion diffusion as more random pathways are accessible within the material. The dependence of the capacity on the Clcontent can be ascribed to increased (local) disorder and formation of “channels” induced by the Ti−Cl in the structure, as the crystal radius of Cl− (167 pm)56 is wider compared to O2− (126 pm).56 Furthermore, the change in local ionic charge distribution (due to the lower electronegativity of Cl) might reduce the Coulombic repulsion, which facilitates Li+-ion diffusion. To determine enhancements in terms of Li+-ion diffusion coefficient as a result of Cl-doping, potential-dependent electrochemical impedance spectroscopy (PEIS) experiments were performed for our best performing Cl-doped am-TiO2 film and the am-TiO2 reference (Figure 4). The Nyquist plots reveal that the overall impedance is smallest for the Cl-doped film, which supports the enhanced rate-performance results. The impedance results were fitted by a small-signal equivalent model based on the one reported by Crain et al. for Li4Ti5O12 (see inset Figure 4b).57 This model consists of a series resistance (Rs), a double layer capacitance (Cdl), a chargetransfer resistance (Rct), a generalized Finite Length Warburg (FLW), and an intercalation capacitance (Cint). The FLW element is given by57,58
results above, the more chlorine in TiO2−xCl2x, the higher the Li+-ion storage capacity, and the maximum capacity at 0.25 C intercepts the capacity of Cl-free am-TiO2 around x = 0.06 Hence, for an enhancement in maximum capacity, a composition of TiO2−xClx with x > 0.06 is required. For x = 0.09 an enhancement of 5 times the capacity of the Cl-free amTiO2 is obtained. At this point, no saturation is yet seen and further enhancement could be possible for even higher Clcontent. To confirm the reproducibility of the electrochemical properties, additional Cl-doped am-TiO2 films deposited at 100 °C with 70 ms and at 115 °C with 140 ms were measured (see Figure S8). At most, a difference in capacity of 4% was measured between two different samples deposited with the same deposition conditions. Gravimetric capacities as a function of C-rate of all samples can be found in Figure S9. These values were obtained by multiplying the volumetric capacity with the RBS density (Table 1). The maximum delithiation capacity of the film with highest Cl content (TiO1.956Cl0.088) has a gravimetric capacity of 362 mAh g−1 at 0.25 C. This value is 8% higher than that expected for insertion of 1 Li per TiO2 (= 336 mAh g−1). An apparent capacity above the theoretical value can come from errors in the measured Ti content from RBS. On the other hand, it is also possible that the excess in capacity is related to the reduction of Ti(III) to Ti(II). Thermodynamically, the reduction of Ti(III) to Ti(II) in crystalline TiO2 for lithium insertion has been calculated by ab initio methods to be energetically favorable at a potential of 0.37 V vs Li+/Li.51,52 More research will be necessary to confirm the formation of Ti(II) states, by, for example, in situ XPS or EELS studies. In any case, the capacity for our best performing Cl-doped electrode reaches the theoretical limit imposed by the Ti(IV) to Ti(III) reduction. In Figure 3, we compare the gravimetric capacity of our best performing chlorine doped am-TiO2 (TiO1.956Cl0.088) and the Cl-free am-TiO2 electrode with nanosized TiO2 electrodes reported in the literature (compiled in Table S1). The gravimetric capacity of our electrode is based on the density determined by RBS (Table 1). To compare the different
Figure 3. Comparison of storage capacities at 1 C (= 336 mA g−1) compiled from literature for TiO2 electrodes. Capacity values are grouped by morphology (nanostructure, carbon composite, and crystal structure) or modification type (i.e., doping) and given as a function of the characteristic length scale. Chlorine free am-TiO2 is deposited from TTIP + H2O and TiO1.956Cl0.088 is deposited at 100 °C from TiCl4 + H2O.
FLW = R d(jωτd)−q tanh(jωτd)q
(5)
with Rd the diffusion resistance, j = −1 , ω the perturbation frequency, and τd the characteristic diffusion time for Li+-ion insertion and extraction in the electrode. For an ideal FLW, q = 10012
DOI: 10.1021/acs.chemmater.7b03478 Chem. Mater. 2017, 29, 10007−10018
Article
Chemistry of Materials
lower across all potentials, with values ranging between 3 × 10−13 and 7 × 10−13 cm2 s−1. This confirms that Cl-doping enhances the Li-ion diffusion coefficient in am-TiO2, resulting in its superior rate performance. 3.4. Demonstration of Cl-Doped am-TiO2 as 3D ThinFilm Electrode. To enable small autonomous devices, integrated energy storage is needed that offers a small footprint and fast charging capabilities. A battery envisioned for this is the three-dimensional (3D) all-solid-state thin-film battery.18,60 Key to its practical implementation are cost-efficient conformal coating techniques and high rate-performance electrode materials. There are several benefits of integrating Cl-doped am-TiO2 thin-films deposited by S-ALD in all-solid-state 3D batteries. First, the relatively low deposition temperature (
