3D aligned-carbon-nanotubes@Li₂FeSiO₄ arrays as high rate capability cathodes for Li-ion batteries

A schematic representation of the synthesis route for the 3D ACNTs@Li₂FeSiO₄/Al samples is shown in figure 1. Aligned-carbon-nanotubes (ACNTs) were synthesized on a circular Al foil (16 mm in diameter) by means of chemical vapor deposition (CVD) with ferrocene (Avocado, 98%) as the catalyst, as has been reported previously [13]. The mass loadings of ACNTs on the Al foil were controlled by the deposition time. Three different ACNTs/Al samples with ACNT mass loadings of 0.5, 1, and 2 mg were prepared. A Li–Fe–Si–PVA solution (molar ratio Li:Fe:Si:PVA = 2:1:1:1.6,[Fe³⁺] = 0.26 M) was prepared by a wet chemical method previously described [6], where Li(CH₃COO)⋅2H₂O (Sigma-Aldrich, reagent grade), Fe(NO₃)₃⋅9H₂O (Sigma-Aldrich, >98%), and tetraethyl orthosilicate (TEOS) (Aldrich, >99%) solutions were used as metal precursors. Polyvinylalcohol (PVA) (Aldrich, Mowiol 10-98, Mw = 61 000) was used as complexing and reducing agents during the carbothermal reaction. The prepared solution (25 μl) was applied drop-wise on the top of the ACNTs on the circular Al foil substrate. This process was followed by aging at 24 ° C for 10 h without covering to obtain a gel-coated composite, and heat-treatment in flowing Ar at 600 ° C for 10 h to obtain the final 3D ACNTs@Li₂FeSiO₄/Al cathodes. Three different 3D samples (mass ratio ACNTs: Li₂FeSiO₄ = 0.5:1, 1:1, and 2:1, respectively) were prepared and are listed in table 1. These samples are denoted as ACNTs0.5@LFS1, ACNTs1@LFS1, and ACNTs2@LFS1 for short in the following.

Zoom In Zoom Out Reset image size

Nanoporous Li₂FeSiO₄/C composites were synthesized by the wet chemical method as previously described [6]. A Li–Fe–Si–PVA solution (molar ratio Li:Fe:Si:PVA = 2:1:1:0.8) was prepared by the same method as described above. The as-prepared solution was stirred at 60 ° C and gel formation took place after stirring for 6 h in an uncovered beaker. The gel was then covered and aged at 24 ° C for 10 h before drying at 130 ° C for 3 h. The Li–Fe–Si-containing dry gels were calcined at 450 ° C for 1 h in air. The calcined powder was then mixed with an aqueous corn starch (Sigma-Aldrich, reagent grade) solution and ground into a paste in an agate mortar. The powder mixtures (starch content = 27 wt%) were heat treated in a flowing N₂ atmosphere at 650 ° C for 10 h to obtain the nanoporous Li₂FeSiO₄/C composite which is denoted as nanoporous sample for short in the following.

The 3D sample ACNTs2@LFS1 with optimized mass ratio of ACNTs and Li₂FeSiO₄ (2:1) and the nanoporous sample were analyzed by x-ray diffraction (XRD) using Cu Kα radiation (Bruker AXS D8 FOCUS diffractometer with a LynxEye PSD). TOPAS R (Bruker AXS) version 2.1 was used for x-ray Rietveld refinements. The specific surface area and pore size distribution were analyzed by nitrogen adsorption measurements (Tristar 3000 Micrometrics). Considering that the surface area of the Al foil can be neglected compared to the ACNTs@Li₂FeSiO₄ nanocomposites, the 3D sample ACNTs2@LFS1 with Al substrate was rolled up and placed inside the glass tube for the nitrogen adsorption measurement. The mass of ACNTs@Li₂FeSiO₄ nanocomposites loading on the Al foil was determined by the weight difference between the ACNTs@Li₂FeSiO₄/Al and the Al. The morphologies of the 3D ACNTs2@LFS1 and nanoporous samples were studied using field emission scanning electron microscopy (FESEM, Hitachi S-4300SE) and transmission electron microscopy (TEM, JEOL 2010F operated at 200 kV). To investigate any changes in the morphology of the samples during cycling, the electrodes were characterized by TEM also after cycling. This was performed by bringing the cathode to a fully discharged state after cycling, followed by disassembling the coin cell, dipping the electrode into a diethyl carbonate (DEC) solution for 10 h to remove any residual electrolyte, and finally drying at room temperature overnight inside an argon-filled glove box. TEM specimens were prepared by embedding the samples in epoxy before cutting thin electron transparent slices with a diamond knife (standard ultra-microtomy TEM sample preparation scheme). The thin slices were further transferred to a holey carbon coated Cu TEM grid. Quantitative chemical composition analysis and elemental mapping were performed with an Oxford instruments energy dispersive spectrometer (EDS) equipped with an X-Max large area (80 mm²) silicon drift detector (SDD). A long acquisition time, typically around 2 h, was used to provide good statistics in the element maps. In addition, drift correction was performed every 30 s during the acquisitions to maintain a high spatial resolution in each of the maps.

The electrochemical performance of the 3D ACNTs@Li₂FeSiO₄/Al and nanoporous Li₂FeSiO₄/C samples was assessed using CR2016 coin cells. The 3D sample could be assembled directly into the coin cell, eliminating the need for binder and conducting additives. However, for the nanoporous Li₂FeSiO₄/C sample, the cathode was prepared by mixing 85 wt% of the nanoporous Li₂FeSiO₄/C with 10 wt% Super-P carbon black and 5 wt% polyvinylidene fluoride (PVDF) (Kynar, reagent grade). Slurries were made by ball milling, using N-methyl-2-pyrrolidone (NMP) (Sigma-Aldrich, >99%) as the solvent. The electrodes were formed by tape casting the slurry onto Al foil, followed by drying overnight at 90 ° C in a vacuum furnace. A typical cathode loading was 4–5 mg cm⁻². Coin cells were assembled with the cathode, lithium metal as the anode and a Celgard 2400 film as the separator. The electrolyte used was 1 M LiPF₆ (Aldrich, ≥99.99%) dissolved in ethylene carbonate (EC, Sigma, 99%)/diethyl carbonate (DEC, Aldrich, ≥99%) (3:7 volume ratio). Cell assembly was carried out in an argon-filled glove box, with water and oxygen concentrations of 0.1 ppm. Charge–discharge analysis was performed galvanostatically between 1.5 and 4.6 V versus Li/Li⁺ with a Desktop Automated Test System (Maccor, Model 4200) at 24 ° C. For the 3D ACNTs@Li₂FeSiO₄/Al sample, the reported capacities are quoted with respect to the Li₂FeSiO₄ loading on the ACNTs, which was determined by the weight difference between ACNTs/Al and the 3D ACNTs@Li₂FeSiO₄/Al, and controlled by the volume of the Li–Fe–Si–PVA solution used for the synthesis. For the nanoporous Li₂FeSiO₄/C sample, the reported capacities are quoted with respect to the mass of the Li₂FeSiO₄/C composite, including the carbon and the secondary phases. Long-term cyclic stability of the ACNTs0.5@LFS1, ACNTs1@LFS1, and ACNTs2@LFS1 was evaluated at a charge/discharge current density of C/20 (C = 160 mA g⁻¹) for the first two cycles and 5 C for the following cycles between 1.5 and 4.6 V versus Li/Li⁺. Over 1000 cycles were performed on these 3D samples.

The electrochemical impedance spectra (EIS) and cyclic voltammetry (CV) of the 3D ACNTs2@LFS1 and nanoporous samples were assessed using a 3-electrode cell (Hohsen Corp.) with metallic Li foil as both reference and counter electrodes. EIS were measured using a multi-channel potentiostat (VMP-3, Princeton Applied Research) in the frequency range of 100 kHz to 10 mHz. The EIS data were simulated by the ZSimpWin software. CV data were collected with VMP-3 at potentials between 1.5 and 4.8 V versus Li⁺/Li and a scanning rate of 0.1 mV s⁻¹.

Figure 2 shows the XRD pattern of the ACNTs@Li₂FeSiO₄ nanocomposites scraped off the Al foil substrate for the 3D sample ACNTs2@LFS1. The broad peaks are attributed to the nanosized particles. The XRD pattern of the nanoporous Li₂FeSiO₄/C composite is also included in figure 2, and the Li₂FeSiO₄ phase could be identified by the space group P2₁/n [14]. Small amounts of Li₂SiO₃ and α-LiFeO_2−x related secondary phases were observed in the nanoporous sample, but the yield of the Li₂FeSiO₄ was as high as ∼90 wt% determined by the Rietveld refinement. Comparing the two XRD patterns, the peak positions of the Li₂FeSiO₄ phase in the 3D sample are corresponding well with the ones for the nanoporous sample. However the unidentified peak appearing at 21° may be due to either a symmetry change of the P2₁/n structure or an impurity. The additional broad peaks appearing at approximately 26° and 42° correspond to the {0 0 2} and {1 0 0} crystal planes of the ACNTs, respectively [15]. A minor (110) diffraction line from Fe can also be observed in the patterns at 45°.

Zoom In Zoom Out Reset image size

The morphology of the bare ACNTs on the Al foil before deposition was investigated by FESEM, as shown in figure 3(a). CNT arrays up to 60 mm long with very clear tops could be obtained by the facile CVD method [13]. The top view of the bare ACNTs shown in the inset of figure 3(a) indicates that the distribution of the CNT arrays was uniform. The FESEM image of the 3D sample ACNTs2@LFS1 viewed from the top, shown in figure 3(b), shows a cornfield-like ACNTs@Li₂FeSiO₄ nanocomposite layer covering the Al foil. The tufted ACNTs bundles could be caused by the capillary forces during drying after deposition of the Li–Fe–Si–PVA solution. A higher magnification FESEM image (figure 3(c)) taken of the cross-section shows the as-prepared ACNTs@Li₂FeSiO₄ nanocomposite with a thickness of approximately 60 μm. The interface part of the nanocomposite adjacent to the Al substrate was investigated by low angle annular dark field (LAADF) scanning TEM (STEM), as shown in figure 3(d). There is a homogeneous and dense coverage of bright nanoparticles on the CNTs. The inset in figure 3(d) showing a higher magnification image of the nanocomposite that is marked by a square in this same figure, clearly indicates that these nanoparticles have a size of around ∼20 nm and are well dispersed on the CNTs. This is marked by yellow lines in the inset. In order to confirm the chemical composition of the nanoparticles, element maps were acquired by EDS on the area defined by the inset in figure 3(d). Fe, Si, O, Al, and C were detected, and the calculated chemical compositions correspond to the Li₂FeSiO₄ phase, as shown in table S1 (available at stacks.iop.org/Nano/24/435703/mmedia). The element maps given in figure 3(e) confirm that most of the deposited nanoparticles are Li₂FeSiO₄. One such particle is marked by a blue circle in the inset of figure 3(d). However, a few iron nanoparticles marked by red circles in the same figure might result from the ferrocene catalyst during the CVD process. A selected area electron diffraction (SAED) pattern of the nanocomposite taken in the upper part of the nanotubes is also included figure 3(e). The two inner diffraction circles correspond to the {0 0 2} and {1 0 0} crystal planes of the ACNTs, which was also observed in the XRD pattern. The diffraction spots appearing in figure 3(e) are mainly attributed to the Li₂FeSiO₄ phase.

Zoom In Zoom Out Reset image size

The LAADF STEM images of the 3D sample ACNTs2@LFS1 after cycling and washing with the DEC solvent are shown in figure 4. After cycling 12 times, the degree of alignment of the nanotubes seemed to be maintained (figure 4(a)). There are no significant morphology changes compared to the uncycled sample (figures 3(b) and (c)), and the main phases were still crystalline, confirmed by SAED. However, large amounts of agglomerates were observed after cycling 50 times (figure 4(b)). The SAED pattern of the nanocomposites (figure 4(c)) only shows the diffraction circles of ACNTs and a few diffraction spots corresponding to bcc Fe, indicating the agglomerates are mainly amorphous. Thus an evolution from crystalline to amorphous phase occurred during cycling. EDS mappings (figure 4(c)) of the area shown in the inset in figure 4(b) show that Fe, Si, O, P, and Al were detected in these agglomerates. Also small amounts of F were present. C is the major element in the EDS spectra. However, since the TEM specimens are resting on an amorphous carbon film, C is omitted from the analysis in order to provide reliable information about the nanoparticles dispersed on the ACNTs. Apart from Fe which appears as crystalline nanoparticles, the elements tended to distribute more homogeneously in the 3D networks for the sample after cycling 50 times and DEC washing compared to the uncycled 3D sample (figure 3(e)). Another interesting observation was that EDS of the sample cycled 50 times after DEC washing showed variations in the Si:O ratio across the cross-section. At the aluminum/cathode interface the Si:O ratio was approximately 1:3 while it was closer to 1:2 at the surface. Hence, the results confirmed that the Li₂FeSiO₄ phase no longer existed.

Zoom In Zoom Out Reset image size

The chemical compositions of the 3D samples ACNTs2@LFS1 before and after cycling and DEC washing are provided in figure 4(d) and table S1 (available at stacks.iop.org/Nano/24/435703/mmedia). The selected areas for EDS analysis are taken at different locations from the surface region to the Al/nanocomposite interface region. The decreasing at.% of Fe with cycling was mainly attributed to the fact that the amorphous silicate phase is more likely to release Fe into the DEC solvent during washing compared to the crystalline Li₂FeSiO₄ phase. The Si:O ratio of the washed samples as well as the at.% of P and F increased after cycling, which are attributed to the products of electrolyte decomposition or active electrode dissolution forming SEI films (e.g. Li_xPF_y, Li_xPO_yF_z and LiF) [16] in the 3D cathode. Despite these morphological and chemical composition changes, the 3D ACNTs@Li₂FeSiO₄ nanocomposites seemed to maintain a favorable pore morphology which allows access of the electrolyte into the network.

FESEM images of the nanoporous Li₂FeSiO₄/C composites before and after cycling are given in figure 5. The surface morphology underwent severe changes after 50 charge/discharge cycles. The nanoporous morphology can no longer be observed clearly, which may be due to the SEI film forming on the surface of the Li₂FeSiO₄/C particles.

Zoom In Zoom Out Reset image size

The BET surface area of the 3D sample ACNTs2@LFS1 and nanoporous sample before cycling were 41 and 20 m² g⁻¹, respectively. The pore size distribution of the two samples, given in figure 6, shows that the nanoporous sample contains mesopores with uniform pore distribution and pore sizes of ∼20–30 nm, while the 3D sample contains a much larger mesopore volume than the nanoporous sample, allowing easier penetration of electrolyte into the cathode particles.

Zoom In Zoom Out Reset image size

Figure 7(a) shows the results of the first and second cycle specific charge/discharge capacities at a current density of C/20 for the 3D ACNTs2@LFS1 and the nanoporous Li₂FeSiO₄/C samples. The 3D sample can deliver a discharge capacity of 175 mAh g⁻¹ during the first and second cycle, while the nanoporous Li₂FeSiO₄/C sample can only deliver a discharge capacity of 80 mAh g⁻¹. Figure 7(b) shows the charge/discharge capacity of the 3D sample ACNTs2@LFS1 at various current densities. The cell shows excellent rate capability, delivering a charge/discharge capacity of 91 mAh g⁻¹ at 20 C (3200 mA g⁻¹), which is 52% of the value obtained at C/20. The discharge capacities of the sample with C rates of 0.5, 1, and 5 were 142, 125, and 112 mAh g⁻¹, respectively. Figure 7(c) gives the cycling stabilities of the 3D sample ACNTs2@LFS1 and the nanoporous Li₂FeSiO₄/C sample at 0.5 C. For the 3D sample, the discharge capacity decreased slightly from 144 mAh g⁻¹ on the first 4–5 cycles, but stabilized after about the fifth cycle and maintained a value of 142 mAh g⁻¹ for the following 45 cycles, which is around 99% of the initial discharge capacity. The nanoporous Li₂FeSiO₄/C showed an initial discharge capacity of 55 mAh g⁻¹, and suffered from about 8% fading after 50 cycles, which is by far inferior to the 3D sample. Furthermore, after cycling both the 3D sample ACNTs2@LFS1 and the nanoporous Li₂FeSiO₄/C sample 50 times, the C rate was increased to 1, 5, and 20 C, and cycled 20 times at each C rate, followed by a return to C/20. The results are presented in figure 7(d). The 3D sample presented good cycling stability at each rate. After a total of 112 cycles (2 cycles at C/20, 50 cycles at 0.5 C and 60 cycles at higher C rates), the reversible capacity at C/20 was 155 mAh g⁻¹, which shows a capacity fading of 12% compared to the initial reversible capacity of 175 mAh g⁻¹ at C/20. Upon further cycling at increased C rates, the reversible capacities were still maintained at 123, 112, and 94 mAh g⁻¹ at rates of 1, 5, and 20 C, respectively. The nanoporous Li₂FeSiO₄/C samples on the other hand, gave much lower reversible capacities (figure 7(d)). Values of 40, 25, and 8 mAh g⁻¹ were measured at cycling rates of 1, 5, and 20 C, respectively. It should be noted that although the capacities were much lower than the 3D sample, they still remained stable for the 20 cycles at each cycling rate. Also, the capacity measured at C/20 after 112 cycles, was close to the initial reversible capacity. Figure 7(e) shows the long-term cycling stabilities of the 3D samples with different mass ratios of ACNTs and Li₂FeSiO₄ at a high charge/discharge current density (5 C). The 3D sample ACNTs2@LFS1 with the highest ACNTs content (66 wt%) delivered the highest initial discharge capacities at both low current density (C/20) and high current density (5 C). The 3D samples ACNTs1@LFS1 (50 wt% ACNTs) and ACNTs0.5@LFS1 (33 wt% ACNTs) gave much lower initial discharge capacities. Values of 145 and 76 mAh g⁻¹ were measured at cycling rates of C/20 and 5 C for ACNTs1@LFS1. For ACNTs0.5@LFS1, the values are 20 and 6.4 mAh g⁻¹. All the 3D samples suffered from about ∼20% fading after 1000 cycles at 5 C. However, the capacity fading during the next 1000 cycles (from 1000 to 2000 cycles), was less severe. Both of the 3D samples—ACNTs2@LFS1 and ACNTs1@LFS1—showed good stability with very little capacity loss from 1000 to 2000 cycles.

Zoom In Zoom Out Reset image size

Electrochemical impedance spectra (EIS) were measured to explore the origin of the improvement in the electrochemical performance of the 3D sample ACNTs2@LFS1 and are presented in figure 8. Nyquist plots of the 3D sample ACNTs2@LFS1 (figure 8(a)) and nanoporous Li₂FeSiO₄/C sample (figure 8(b)) measured before charging and at fully discharged states at different cycles consist of a small semicircle in the high-frequency region (although not very clear in figure 8(a)), and a large semicircle in the high-to-middle frequency region, which are attributed to the lithium-ion migration through the SEI film and the charge transfer through the electrode/electrolyte interface, respectively [17]. However, some major differences can be observed in figures 8(a) and (b). In addition to the obvious fact that the total impedance of the 3D sample is much lower than the nanoporous sample, there are also large deviations in the low frequency region. The nanoporous Li₂FeSiO₄/C sample shows a straight line with a 45° angle, corresponding to a semi-infinite Warburg diffusion process in the bulk [18]. The 3D sample ACNTs2@LFS1 on the other hand, shows a line with an angle much larger than 45° and approaches 90° in the lower frequency region, corresponding to a transition region and finite diffusion region [19]. This indicates that Li-ion diffusion was sufficiently fast due to the short diffusion length. The semi-infinite diffusion controlled region was not observed and a capacitive effect was emphasized on the Nyquist plot due to charge saturation [19] of the Li₂FeSiO₄ nanoparticles or Li–Fe–Si–O amorphous phase.

Zoom In Zoom Out Reset image size

Based on the above analysis, simulated electrochemical parameters are shown in tables S2 and S3 in the supplementary materials (available at stacks.iop.org/Nano/24/435703/mmedia) by using the equivalent circuit in figure 8(c). This equivalent circuit proposed by Meyers et al [20] is widely used for simulation of porous electrodes. Here, R_e,R_SEI, and R_ct are resistances of the electrolyte, the SEI film, and the charge transfer reaction, respectively. The capacitance of the SEI film and the double layer are represented by the constant phase elements (CPE) Q_SEI and Q_dl, respectively, due to the 'depressed semicircle' in the impedance spectra [17]. The low frequency region cannot be modeled properly by a finite Warburg element. Therefore, the constant phase element (CPE), i.e., Q_D was also used to replace the diffusion element. This approach was used to obtain a good agreement between the calculated and the experimental data, as shown in figures 8(a) and (b). The significant change of the Nyquist plots for different samples during the cycling is the values of R_SEI and R_ct, as listed in table 2. The R_SEI and R_ct of the 3D sample are much lower than that of the nanoporous Li₂FeSiO₄/C sample both before and after the same number of cycles. In addition, the R_SEI and R_ct increased after cycling for both the 3D and the nanoporous Li₂FeSiO₄/C samples. This increase might be caused by the formation of an SEI film, electrode amorphization and dissolution (Fe loss during cycling) and would consequently result in the observed capacity fading after 50 cycles (figure 7(c)), especially the 113th cycle at low current density (C/20) (figure 7(d)).

The values of the exchange current density (j₀) are calculated using equation (1) [21]:

$$\begin{eqnarray} &&{j}_{0}=\frac{R T}{n F{R}_{\mathrm{ct}}A} \end{eqnarray} \tag{ 1 }$$

where R is the ideal gas constant, T is the temperature (297.5 K), F is the Faraday constant, A is the area of the electrode surface (2 cm²) and n is the number of electrons transferred per ion during the intercalation, which is 1 for Li⁺. The exchange current density is a parameter that can be used to indicate the reversibility of the electrode. The higher value of j₀ observed for the 3D sample implies better reversibility compared to the nanoporous sample. The above results are also consistent with the cyclic voltammetry analysis presented in figures 9(a) and (b) for the 3D sample ACNTs2@LFS1 and the nanoporous Li₂FeSiO₄/C sample. In the initial anodic process, oxidation peaks around 3.1 V versus Li⁺/Li are evident in both samples, which can be attributed to the oxidation of Fe²⁺ to Fe³⁺ [1]. After cycling, the Fe²⁺/Fe³⁺ redox potential dropped about 0.3 V, which was mainly due to the P2₁/n polymorph of Li₂FeSiO₄ transforming into the more stable inverse β_II (Pmn2₁) polymorph during initial charging [22]. After the phase transformation, the Pmn2₁ phase was reversibly charged and discharged without further phase transformations. For the 12th cycle, the peak profile of the 3D sample was more symmetric and sharp, indicating a smaller conductivity restriction than that of the nanoporous sample. The redox potential separation of the 3D sample was 0.31 V, which is lower than that of the nanoporous sample (0.47 V). For the 50th cycle, the redox potential separation of the 3D sample remained low and the peak profile was still sharp. Additional oxidation and reduction peaks appearing around 3.2 V versus Li⁺/Li indicate the charge/discharge mechanism changed from the first few cycles. However, the redox potential separation of the nanoporous sample increased upon cycling and the peak profile became broader than the nanoporous sample after 12 cycles. The low redox potential separation for the 3D sample implies enhanced electrode reversibility and weaker polarization compared to the nanoporous sample. In addition to the oxidation peak at ∼3.1 V, another peak around ∼4.8 V is also present in the CV profile, especially for the 3D sample. This peak could be attributed to the Fe³⁺/Fe⁴⁺ redox couple, which forms in the voltage of 4.1–4.8 V [23]. Thus, more than one Li⁺ (e.g., 1.05 mol per formula unit) could be extracted out of/inserted into the Li₂FeSiO₄ lattice at a rate of C/20 (figure 7(a)). It should be noted that the CV peak profiles at high voltage (>4.5 V) may also be attributed to the decomposition of bulk electrolyte components [24]. Further work is needed to verify this.

Zoom In Zoom Out Reset image size