Performance of Negative Lithium Titanate Electrodes Containing Minimized Amounts of Conducting Polymer and Modified Guar Gum as Binder

Li₄Ti₅O₁₂ (LTO) powder (< 200 nm) and PEDOT:PSS 1.3 wt% aqueous dispersion were purchased from Aldrich, carbon black «Super P» (CB) from Timcal Inc. (Belgium); commercial battery electrolyte solution TCE-918 from Tinci Materials Technology Co. Ltd. (China). Polyflos HPG15 (Hydroxypropyl guar) was provided by Lamberti S.p.A. LTO powder was dried at 120°C under vacuum. All others reagents were used as received.

Components were mixed at a ratio 90 wt% of LTO, 6 wt% of CB, 2 wt% of HPG15 and 2 wt% of PEDOT:PSS. Firstly, HPG15 and PEDOT:PSS dispersion were mixed and a small amount of deionized water was added. Into this viscous solution LTO powder was added, and the mixture was blended for 15–20 minutes. Finally, CB was added and the slurry was mechanically mixed 1 hour until the mixture became homogeneous. The resulting slurry was cast onto Al foil used as support and current collector using a doctor blade applicator set at 150 μm thickness and dried at room temperature in air atmosphere for 10 hours. All electrodes were cut into disks with 1.77 cm² area and average mass of electrode material of about of 5 mg.

LTO-electrodes were tested in standard coin-type cells CR2032. The cells were assembled in an argon-filled glove box Unilab (USA) using Celgard 2325 membrane as separator, Li foil as counter electrode, and the battery electrolyte solution TCE-918. All electrode potentials are reported with respect to Li/Li⁺. This implies that the studied electrode intended for use as a negative electrode in a lithium ion battery was the positive electrode in our experiments. To avoid confusion terminology (lithiation/delithiation) always pertains to the intended use.

The electrochemical performance tests were carried out on an automatic galvanostatic charge-discharge cell tester (Neware Co., China) in the potential range between 1.0 V and 2.5 V (vs Li/Li⁺) at different discharge rates (from 0.2 C to 50 C) (current 1 C is equal to 175 mA·g⁻¹) and constant charge rate (0.2 C) at room temperature (22°C). Theoretical specific capacity of LTO was calculated as 175 mAh·g⁻¹. All capacity values were normalized to the total mass of the electrode.

Cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) were performed with an Autolab PGSTAT 30 potentiostat/galvanostat (Eco-Chemie, Netherlands) equipped with FRA2 module using the button cells described above, i.e. in 2-electrode arrangements. Cyclic voltammograms were recorded in the potential range 1.0–2.5 V vs Li/Li⁺ at a scan rate 0.1 mV·s⁻¹. Impedance measurements were performed over a frequency range of 100 kHz to 0.1 Hz with an applied amplitude of 5 mV. Impedances were measured in the fully charged state (at E = 1.0 V), fully discharged state (E = 2.5 V) and in state of charge (SOC) 50 % (E = 1.55 V) after electrochemical performance tests. To achieve a steady state, cells were conditioned by keeping them for 2 hours at indicated potentials after 3 CV cycles. To examine changes in electrode kinetics after extended cycling impedances were measured before and after 200 charge/discharge cycles. The parameters of the equivalent circuits were calculated and analyzed using Nova 1.11 software.

Fig. 1 shows charge/discharge plots for a cell with an LTO electrode containing the mixed HPG-binder at various cycle numbers, for comparison curves for a cell with PVDF binder are shown also.

Zoom In Zoom Out Reset image size

The curves illustrate the remarkable stability of this electrode material; actually, the trace of the 100^th cycle shows toward the end of the cycle a slightly better performance than in previous cycles. As discussed elsewhere¹⁸ the inferior stability of an electrode containing PVDF as a binder is visible already in the 100^th cycle, more details are discussed below. Polarization in terms of the difference between the upper charge and the lower discharge traces being due to both Ohmic as well as kinetic contributions hardly changes between the first and the 100^th cycle for the LTO with Polyflos. There is also no decrease of discharge capacity. Quite differently an increase of polarization and a decrease of capacitance is observed after 100 cycles with LTO and PVDF as a binder.

Rate performance is demonstrated in Figs. 2a and 2b.

Zoom In Zoom Out Reset image size

Characteristic charge and discharge plateaus are observed at low current densities up to 1 C, at higher rates polarization visibly increases until 10 C (see in particular Fig. 2b). Reversibility especially after high rate discharge is notable. Specific capacity values at low current densities reached 149 mAh·g⁻¹ at 0.2 C and decreased to 107 mAh·g⁻¹ at 10 C. After measurements at 10 C extended cycling at 1 C was performed. Fig. 3 shows that capacity values thereafter returned to the values observed before at 0.2 C. Capacity values are very stable, at least up to 100 cycles no capacity loss is observed. Actually at low rate (0.2 C) a slight improvement is observed, for more details see Fig. 4.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The benefits of using a biopolymer (CMC) combined with PEDOT:PSS instead of PVDF as a binder alone in both positive and/or negative electrodes in terms of various performance parameters have been specified elsewhere already.^18,20

Further benefits gained possibly by substituting CMC with HPG15 may become obvious when comparing long-term cycling data for representative samples as shown in Fig. 4.

Obviously in terms of stability, the binder based on HPG15 and PEDOT:PSS works better than PVDF, but specific capacity of this electrode is slightly lower than that with a PEDOT:PSS/CMC binder. Stability expressed in terms of capacity loss per cycle is 2.19·10⁻³ mAh·g⁻¹ for the CMC-containing material,¹⁸ 8.80·10⁻² mAh·g⁻¹ for the PVDF-containing material and -8.75·10⁻³ mAh·g⁻¹ for the material examined here based on initial capacitance and the respective value in the 100^th cycle. Because the initial break-in period and some data scattering make a simple comparison end value vs. start value in particular with the present electrode material unreliable interpolation was applied, the result was -3.3·10⁻³ mAh·g⁻¹. For the CMC-containing material, the decrease obtained this way is 4.2·10⁻³ mAh·g⁻¹. The capacity decrease of the PVDF-containing material is non-linear. Both comparisons show a slightly better stability of the material investigated here. Even after 200 cycles HPG15 works better, the capacity loss is only -1.8·10⁻³ mAh·g⁻¹. This slightly higher stability is possibly related to the hydroxylation of modified guar gum used here providing more efficient interaction between biopolymer and LTO particles.

This intermediate performance can also be observed in rate testing. As shown in Fig. 5 rate retention of the present electrode material at higher C rates is between the retention observed with LTO containing CMC-based binder and PVDF as a binder. The PVDF-bound LTO electrode showed the worst high-rate capability. The specific capacity of this electrode was approaching zero at a current density equivalent to 30 C. This is mostly related to the lower conductivity of LTO electrodes with PVDF binder. Elsewhere capacity values LTO grain material increased after carbon coating up to 160 mAh·g⁻¹ (at 0.1 C) and 75 mAh·g⁻¹ (at 1 C) in comparison with pristine LTO material.²¹ For LTO_{PEDOT:PSS/HPG15}-electrode studied here a better C-rate capability was achieved when compared with,²¹ specific capacity values are 149 mAh·g⁻¹(at 0.2 C) and 148 mAh·g⁻¹ (at 1 C).

Zoom In Zoom Out Reset image size

To improve understanding of the performance of the electrode material cyclic voltammetry and impedance measurements were performed.

Cyclic voltammograms of the LTO-electrode as a function of cycle number are displayed in Fig. 6; for comparison, a CV obtained with an electrode containing PVDF as a binder is shown also.

Zoom In Zoom Out Reset image size

A wide anodic peak is observed, the cathodic peak changed slightly in the initial cycles; in the fifth cycle, a stable behavior was established. Transformed charges were Q_an = 2.68·10⁻⁴ C (Capacity 99 mAh·g⁻¹), Q_cath = 2.74·10⁻⁴ C (Capacity 101 mAh·g⁻¹) indicating high electrode reversibility. The sluggish behavior during delithiation is at variance with the sometimes suggested fast delithiation claimed elsewhere (e.g. in Ref. 22 based on Ref. 23). This asymmetry has been observed with a conventional PVDF-bound LTO electrode before and has been reported elsewhere,^18,20,22 whereas in other reports highly symmetrical peaks have been observed (see e.g.^24,25). Similar wide peak shapes but for both processes at low temperatures also with an electrode of LTO with PVDF as a binder have been reported elsewhere.²⁶ At higher scan rates, i.e. 1 mV·s⁻¹ instead of 0.1 mV·s⁻¹, onset of such asymmetry has been seen also.²⁴ It can also be found with poorly conducting LTO samples without carbon coating or added carbon.^27,28 The asymmetric voltammetric response noticed here in the CV traces indicates differences of charge transfer kinetics during oxidation and reduction, i.e. delithiation and lithiation. It is significantly different from the anodic peak in the CV obtained with CMC or hydroxypropylcellulose as the biopolymeric binder component instead of HPG, which shows a pronounced, narrow positive-going current peak.^20,29 Such highly symmetric behavior has also been observed when using only PEDOT:PSS as binder.³⁰ The asymmetry, in particular the sluggish delithiation, has been examined in detail by Chiu et al.²² In addition to a broader and lower anodic peak when compared to the negative one Tafel-plots of the negative and positive branches of the current-potential curve clearly support the evidence from CV indicating slower delithiation. Reasons were not suggested. Closer examination of²³ reveals that in this report faster diffusion was suggested as reason of the superior rate performance of LTO, not faster charge transfer.

To further examine these differences impedance measurements were made in 2-electrode arrangement (coin cell). Accordingly, cell impedances, not electrode impedances, were obtained.³¹ Interpretation of measured impedances is commonly based on the use of equivalent circuits,³¹ the use of transfer functions appears to be less common.^32,33 The difference between cell and electrode impedances had been addressed in an earlier report already,³⁴ for a more recent example see.³⁵ In that study of lithium primary cells with SO₂ and SOCl₂ as reactants at highly porous carbon positive electrodes it was suggested, that the large surface area of the positive electrodes resulted in a correspondingly small impedance which could be neglected in comparison to the impedance of the lithium electrode. The impedance of various electrodes prepared with different poythiophenes has been considered elsewhere.³⁶ The basic cell impedance of a full (two-electrode) cell as studied before³⁴ and examined here is shown in Fig. 7.

Zoom In Zoom Out Reset image size

If the measured impedance shows evidence indicating the dominance of one electrode impedance by e.g. showing only one time constant as in Ref. 34 a simplification as shown in Fig. 8 may be justified.

Zoom In Zoom Out Reset image size

This circuit commonly known as Randles circuit³¹ has been employed in studies of electrode materials similar to the one examined here before. Bai et al.²⁸ and Liu et al.³⁰ have reported on impedance measurements performed with two-electrode coin cells. In the discussion only electrode impedances are considered, the obvious discrepancy between measured two-electrode cell impedances and discussed electrode impedances is not even mentioned. A visual inspection of the displayed impedances with the only difference between the displayed ones being the degree of fluorination of the titanate used in the electrode suggests, that any conceivable contribution from the lithium counter electrode (this electrode is nowhere mentioned in the report, but its use can be inferred from further details in the text) is constant, at least does not cause a contribution with a second time constant visible in the displayed data. The same equivalent circuit was used in the interpretation of cell impedances with a working electrode made of lithium titanate/carbon nanotube composites by Chen et al.³⁷ Although the displayed impedance data suggest a significant and well visible deviation from the simple shape observed with a Randles-type electrode impedance neither the possibility of a further process causing an impedance feature indicative of a second time constant nor the obvious difference between cell and electrode impedance were addressed. Some authors report data on electrode kinetics derived from impedance measurements without even specifying the type of equivalent circuit or the mode of evaluation (by e.g. a transfer function instead of an equivalent circuit).^20,24,29 Once impedance measurements were just mentioned but not addressed further.²³

A more complicated equivalent circuit depicted in Fig. 9 has been used in a study of lithium titanate negative electrodes with a water-based binder of polyvinyl alcohol and sodium alginate.³⁸ Although measurements were done in a two-electrode arrangement with a lithium reference electrode (presumably, this was also the counter electrode in the Swedgelock-type cell) the discussion presumes measurement of an electrode impedance without providing any arguments for this simplification.

Zoom In Zoom Out Reset image size

The strongly depressed shape of the semicircle-like feature at intermediate-to-high frequencies in the Nyquist plot seems to support the equivalent circuit featuring at least two time constants, a closer discussion would require numerical results of the fit obtained for the film resistance R_f and the film capacitance C_f not reported in the work. This equivalent circuit was also used by Yuan et al.²⁶ albeit with different interpretation of R_f and C_f. The former was called surface polarization resistance, the latter surface capacitance without providing any details and rationale.

An even more complicated equivalent circuit as shown in Fig. 10 was used by Kawade et al.³⁹

Zoom In Zoom Out Reset image size

Resistive elements are solution resistance R_sol, resistance R_L of the silicate coating applied to the titanate particles, contact resistance between the particles R_c and charge transfer resistance R_ct. A seemingly similar equivalent circuit depicted in Fig. 11 has been proposed by Dollé et al.⁴⁰

Zoom In Zoom Out Reset image size

A similar circuit omitting diffusion and with slightly different interpretation of the circuit elements has been used by Kulova et al.³⁵

A more extended circuit for a full cell (see Fig. 12) with lithium and lithium titanate electrodes has been suggested by Schweikert et al.⁴¹ with the inductance of wires etc. L, the solution and electronic resistance R_sol, the interfacial impedance R_cc/Q_cc due to the contact lithium titanate/current collector, the impedance due to the solid electrolyte interface R_SEI/Q_SEI, the charge transfer impedance at the titanate/electrolyte solution interface R_CT/Q_CT, the diffusion impedance R_Diff/Q_Diff due to lithium ion diffusion in the titanate and an impedance element R_Depl/Q_Depl associated with depletion of lithium ions within the titanate. In a related study Schweikert et al. have inspected symmetrical Li/Li and LTO/LTO cells as well as a Li/LTO cell.⁴² The applied equivalent circuit is the same as shown in Fig. 12 except for different labels at the circuit elements. When comparing the impedance of the two symmetrical cells the semi-arc found with the LTO/LTO cell is substantially smaller than with the Li/Li cell suggesting strongly that the impedance of the Li-electrode should not be ignored simply. The displayed Nyquist-plot found with a Li/LTO cell shows a semi-arc roughly equivalent to half the impedance values of the symmetrical cells when added, exactly as expected when both electrode impedances have been extracted from measurement data obtained with a symmetrical cell as suggested elsewhere³¹ in agreement with considerations provided earlier.³⁴ The assumed additivity of electrode impedances determined by experiments with symmetric cells toward cell impedances has been confirmed by Keil.⁴³

Zoom In Zoom Out Reset image size

The need for three-electrode measurements in studies of lithium ion batteries of various chemistries has been pointed out already by Orsini et al.⁴⁴ Experimental details possibly affecting results and their interpretation have been discussed by Levi et al.⁴⁵ The equivalent circuit proposed in their report is a simplified version of the one shown in Fig. 7. The authors found only one diffusion process, which they assigned to the positive electrode without connecting this element with the respective interfacial capacitance and resistance in the common way. The assignment of these resistive elements as electrode resistance is somewhat unusual. Assuming the common interpretation as charge transfer resistance the provided explanation of the observed changes with cycling is conclusive. The implicit assumptions needed when interpreting the impedance data of two-electrode cells could be justified in subsequent three-electrode experiments using a piece of lithium foil attached to a copper grid as reference electrode. In a subsequent study with lithium titanate as the positive electrode material a reference electrode made of this material partially lithiated was identified a particularly suitable reference electrode.⁴⁰ Elsewhere a lithium-tin reference electrode has been proposed.⁴⁶

From this brief overview of published reports on impedance measurements with LTO electrodes it appears that, there is no generally accepted equivalent circuit. Consequently, we have inspected our results obtained after moderate (C-rate measurements, see Figs. 2 and 5) and extensive cycling (see Fig. 3) at various electrode potentials, i.e. states of charge, as shown in typical examples in Figs. 13 and 14.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Assuming an unknown but constant contribution from the lithium electrode used as counter and reference electrode which changes from anode to cathode but stays always as the negative electrode and will presumably change its behavior not significantly we have compared these results with the published ones discussed above in search for an equivalent circuit as simple as possible providing a satisfactory fit. Fit data obtained with the circuit displayed in Fig. 9 for a cell after rate testing but before extensive cycling are listed following in Table I.

Upon delithiation, R_ct significantly grows to five times its value in the lithiated state. In the absence of data of electrochemically active surface area this might just indicate a larger active area at constant rate of charge transfer or an enhanced charge transfer at constant area.⁴⁷ In any case, this will contribute to a growth of the sum of all Ohmic components of the impedance, i.e. the polarization resistance R_pol.³² R_pol is equivalent to the slope of the current density vs. electrode potential curve at the electrode potential where the impedance measurement was taken (here at constant voltage, i.e. at constant electrode potential of the LTO-electrode vs. Li/Li⁺).^32,48 Growth of the diffusion impedance (note that σ is the inverse of the diffusion resistance³²) and of the film resistance add to this. An increase of R_pol indicates slower electrode kinetics, exactly as observed with the CVs discussed above already.

Impedance measurements performed again with the LTO-electrode¹ in the lithiated and the delithiated state before and after 200 charge/discharge cycles were evaluated with regard to conceivable changes in electrode kinetics. Resulting data are shown in Fig. 14; data obtained by fitting are collected in Table II.

In the lithiated state the most striking change is a significant decrease of R_ct to less than a quarter of its initial value. As indicated above in the absence of data of electrochemically active surface area this might just indicate a larger active area at constant rate of charge transfer or an enhanced charge transfer at constant area. Increases of film resistance and capacitance are also significant.

CVs run with the cell after 200 cycles shown in Fig. 15 support the observation from the impedance measurements: The grown value of the film resistance R_f shows up as a small shift of the lithiation current peak to more negative values than before cycling.

Zoom In Zoom Out Reset image size

Further confirmation of this interpretation of experimental observations can be obtained from GCD-measurements at different C-rates. Results are displayed in Fig. 16.

Zoom In Zoom Out Reset image size

The Ohmic components grown after 200 cycles show up as decreased rate capability (see Fig. 16a) as well as increased polarization in the charge-discharge curves (see Fig. 16b).

An LTO-electrode with a mixed water-based binder combining PEDOT:PSS and hydroxypropylated guar gum avoiding PVDF as the conventional binder shows a specific capacity of 149 mAh·g⁻¹ at 0.2 C with a moderate decrease to 107 mAh·g⁻¹ at 10 C. These values as well as the capacity retention at increasing current densities place this material between previously reported ones employing carboxymethyl cellulose and PVDF as binder components. Stability of the material reported here is slightly better than that of the CMC-containing one and much better than that of LTO-based electrodes containing PVDF as a binder.

Generous donations of samples of Polyflos by Lamberti S.p.A. and financial support from Saint Petersburg State University (grant № 26455158) are gratefully acknowledged.

R. Holze 0000-0002-3516-1918