Carbon nanotube tissue as anode current collector for flexible Li-ion batteries—Understanding the controlling parameters influencing the electrochemical performance

The constant technological advancement has generated a new class of soft and flexible electronics, including energy harvesting devices (e.g., solar-cells), memory devices, biomedical devices, wearable electronics, and sensors.^1,2 Integration of a flexible power source, which is capable of operating even under mechanical deformation, into this type of devices is an indispensable step toward making them truly free from any mechanical restraints. Current research concerning flexible batteries focuses on the development and study of diverse thin-film materials, including polymers, inorganics, and graphene derivatives,^3–6 and their implementation into a battery.

Rechargeable Li-ion batteries are the most common apparatus for energy storage in innumerable portable devices. Their performance is limited by the energy density they can provide, prompting the research toward innovations in the field.^7,8 The conventional assembly of a Li-ion battery is two electrodes isolated by a separator in a cell containing the electrolyte. The anode comprises carbonaceous active material (e.g., graphite) coated onto a current collector, typically Cu, and the cathode comprises a Li metal oxide coated onto an aluminum (Al) current collector. During charging of the battery, Li ions are transported to the carbon anode, where they are intercalated in the spaces between neighboring graphite sheets. Upon discharge, the stored Li is de-intercalated, and the battery provides electricity. The first stage of charging the battery involves the formation of a solid electrolyte interphase (SEI)⁹ on the carbon anode at potentials around 0.5-1 V _{vs. Li/Li}⁺, depending on the nature of the carbon material.¹⁰ This insulating layer allows transportation of Li through it and does not interfere with its intercalation within the anode. It is mainly composed of Li salts: Li oxide (Li₂O), Li carbonate (Li₂CO₃), and Li fluoride (LiF).¹¹ The formation of the SEI is crucial for prevention of Li plating onto the carbon anode during discharge; however, it is accompanied by an irreversible capacity loss,¹² restraining the anode capacity from reaching its theoretical limit (372 mAh/gr).

The current collector material is one parameter significantly influencing the overall performance of the battery. Ideally, current collectors should be inert and have no impact on the overall electrochemical performance of the battery. The metals traditionally used as current collectors in Li-ion batteries, and most particularly Cu, can be designed to be thin and flexible, and the performance of the battery in this case is mainly governed by the choice of active materials. However, metallic current collectors have several disadvantages despite their widespread use. They suffer from electromigration that shortens the battery lifetime and deteriorates its performance, they are rather heavy and increase the total weight of the device, and although the current collectors themselves may be flexible, they enforce certain rigidity on the whole structure since detachment of the active material may become an issue when the electrode is subjected to mechanical deformations.¹³ 

One approach for developing thin and flexible Li-ion batteries is the design of new free-standing electrodes, in which the active material holds both sufficient conductivity and mechanical stability; thus, the current collector becomes obsolete. Such studies are mostly focused on the synthesis of nano-architectures of distinct materials.^14–16 Yet, in many cases, particularly when one wishes to use the existing commercial active materials, the use of a current collector is inevitable.

Two-layered current collectors consisting of a carbonaceous substrate and a thin metal layer may allow exploiting the advantages of a metallic current collector, while the substrate provides the desired flexibility. Such current collectors have shown electrochemical performance that is independent of the current collector.^17,18 Nonetheless, if the metal component remains, the need and demand for an overall weight decrease still exist.

Carbonaceous current collectors offer the opportunity to decrease the electrode weight and increase its flexibility. Conductive polymers, such as polypyrrole and polyaniline, are characterized with discharge capacities not much above 100 mAh/gr and have been investigated as electrode materials for batteries and supercapacitors, with and without additional active material.^5,19,20 Another flexible light-weight replacement for the Cu anode current collector is the carbon nanotube (CNT) tissue.^21,22 Using bare CNT as the active material has been attempted but was found to be a less favorable option due to extensive irreversible capacity loss during the first cycle (∼10³ mAh/gr).^21–26 Graphene anodes may provide a higher reversible capacity than CNT albeit the irreversible capacity loss during the first cycle is still of the same order of magnitude.^27,28 This capacity loss has been attributed to the immense active surface area of the CNT, which increases the amount of Li invested in the formation of the SEI. Moreover, the intercalation of Li can be either on the outer surface of the nanotubes, or in their inner core. Upon de-intercalation (discharge) from inside the nanotubes, Li diffuses through the nanotubes from one end to another or through defects in the crystalline structure. When the length for diffusion is too long (close-end nanotubes, or a defect-free structure), Li is irreversibly intercalated in the inner core of the nanotubes, further exacerbating the irreversible capacity loss.^21,29 Applying a layer of active material onto the CNT fabric has substantially diminished this loss. Nevertheless, a complementary pretreatment to the fabric with either isopropyl alcohol (IPA) or tertbutyl alcohol (tBuOH) was still essential. The particular molecular structure of these alcohols enables stronger and more organized van der Waals interactions between adjacent nanotubes, enforcing them to stay closer together. This effect was fundamental for inducing bundling of the nanotubes, improving tissue’s conductivity, and comparing the performance of an anode utilizing CNT current collector to that of an anode utilizing Cu current collector.^17,24 Nevertheless, the efficacy of the pretreatment depends on the exact characteristics of the tissue, stemming from its manufacturing process. The specific manufacturing process of the CNT tissue changes from one manufacturer to the other, as well as within the natural evolution of more efficient manufacturing processes. Thus, while chemically induced bundling of the nanotubes has already been shown to improve the anode performance,^17,24 a change in the CNT manufacturing process may hinder an efficient bundling, posing a challenge of an uncontrolled process.

This research aims to study the influence of diverse parameters on the anode performance when using CNT tissue as a current collector. The performance of the anode was examined as a function of loaded mass of active material, suggesting an improvement at higher loadings. In addition, alternative pretreatments to the tissue were explored, guided by two main rationales; first, restoring the fruitful bundling of the nanotubes in another manner and second, removing any residual contaminants with a proper cleaning procedure, thus eliminating the need for other alterations in the formerly recommended pretreatment. The implementation of CNT tissue as an anode current collector, using these methods, is expected to further progress the development of flexible, light-weight Li-ion batteries.

In this work, tissues composed of a random network of multi-wall CNT (MWCNT, Tortech Nano Fibers, 0.3 mg/cm²) were used. They were immersed for 1 h in one of the following solutions: IPA; 5% and 10% poly-vinylidene difluoride (PVDF) in N-methyl-2-pyrrolidone (NMP); 10%, 15%, and 30% hydrogen peroxide (H₂O₂); or 10%, 20%, and 40% sulfuric acid (H₂SO₄).

Surface morphology after each pretreatment was assessed by using a high-resolution scanning electron microscope (HR-SEM, Zeiss Ultra-Plus FEG-SEM), equipped with an x-ray energy dispersive spectroscopy (EDS) detector. Raman spectra were recorded with the XploRA Raman spectrometer (Horiba Scientific), equipped with a 532 nm laser.

Slurries of the active material were prepared by dissolving 90% graphite powder (Samsung) with 10% PVDF in NMP. The slurry was spread on CNT tissue using Doctor-blade, followed by drying overnight at air atmosphere, and then under vacuum at 120 °C. Finally, the anodes were cut by a 12.5 mm die punch, providing electrodes with active material loading of 4-20 mg and thickness of 40-170 μm, respectively.

The performance of the anodes was tested in a T-cell configuration. The T-cells were assembled inside an argon filled glove-box, using graphite loaded on a CNT current collector as a working electrode and Li foil as a counter electrode. The electrolyte composition was 1M solution of lithium hexafluorophosphate (LiPF₆) in 1:1 ethylene carbonate (EC) and dimethyl carbonate (DMC) (Sigma Aldrich). Charge-discharge cycles were conducted at room temperature applying 0.1 mA/cm², using the Arbin BT2000 battery test system.

Pristine CNT tissues were loaded with varying mass of graphite active material and used as a working electrode in a half-cell vs. Li counter electrode. The first charge-discharge profiles are displayed in Fig. 1(a) along with a profile utilizing the Cu current collector for comparison. All the charge curves exhibit a step at 0.8 V, indicating the SEI formation. A plateau in the potential can be observed, followed by a gradual decrease until 0.2 V. When utilizing a Cu current collector, this decrease is relatively steep, reaching a capacity of only ∼50 mAh/gr. However, when utilizing a CNT current collector, this potential decrease is more moderate. If no graphite is loaded, this plateau and subsequent potential decrease are very wide, obtaining a capacity of almost 2000 mAh/gr, which is irreversible [Fig. 1(a), inset]. The graphite loading onto the CNT significantly decreases this loss, depending on the exact mass of active material loaded, as can be observed in Fig. 1(b). At the highest loading (20 mg), the step resembles that of a Cu current collector, achieving a similar capacity (50 mAh/gr). However, it becomes wider as the loading decreases, reaching 200 mAh/gr for the 4 mg loading and indicating an extensive capacity aimed and utilized toward SEI building. In agreement, the irreversible capacity loss is also lower at higher loadings. The layer of active material forms a physical barrier between the CNT tissue and the electrolyte. The electrolyte must first diffuse through this layer in order to reach the underlayer CNT, thus inducing the SEI formation on the graphite as well as on the CNT. As more graphite is loaded, the layer is thicker, and the diffusion length for the electrolyte increases, resulting in a lower Li concentration within the CNT. Therefore, the active surface area available for SEI formation is smaller and the irreversible capacity loss it induces decreases as well. This effect is more pronounced in a lower active graphite loading and diminishes above 15 mg.

Figure 1(c) displays the capacity measured during SEI formation in the first cycle as a function of graphite mass. This behavior is proportional to the error function,

The capacity recorded during the first lithiation in the potential range of 0.8-0.2 V is solely attributed to SEI formation. The limiting factor to its formation within the CNT part of the electrode (instead of the graphite) is the concentration of Li ions in it. As this concentration is higher, more extensive SEI formation is possible, as can be seen in the extreme case of a bare CNT anode.

This system can be considered as a semi-infinite diffusion couple, where the thickness of the graphite layer determines the length through which Li ions have to diffuse until reaching the CNT. The initial time when SEI formation begins is equal in all cases, and therefore its effect over the initial concentration of Li ions on the surface can be disregarded. The concentration profile in this case is proportional to the error function [Eq. (1)], coinciding with the experimental results. Therefore, it can be construed that indeed applying a graphite layer onto the CNT poses a diffusion barrier to Li, thus decreasing the capacity loss induced by the presence of CNT.

When operating with high loadings of active material, no additional pretreatment is required; a comparison of the performance of the CNT current collector to that of Cu shows no difference. Yet, since the mass of graphite loaded onto Cu is lower (10 mg), it diminishes the advantage of using a lighter current collector. Furthermore, lower loadings may be of interest, especially for applications of high power, where the light weight is mandatory. Therefore, CNT tissues were pretreated according to the methods described previously and used as anode current collectors with 8 mg of active material loaded.

CNT tissues were pretreated with IPA or a PVDF/NMP mixture, and their surface morphology was characterized. Areas with different densities are observed on the same CNT tissue prior to any pretreatment [Fig. 2(a)], and IPA wash does not seem to considerably change that [Fig. 2(b)]. On the contrary, the pretreatment with PVDF solution resulted in larger percent of bundled areas [Figs. 2(c) and 2(d)]. When the concentration of PVDF is low (5%), only a minor impact is detected. Nevertheless, at a higher concentration (10%), the densification is more effective. Pretreatment with even higher PVDF concentrations is proven to be inefficient, due to the high viscosity of the mixture, which formed lumps, and uneven distribution of PVDF on the surface of the CNT. In the permeation of the PVDF/NMP mixture (also used in the preparation of the slurry) into the tissue, PVDF acts as a gluing agent thus inducing the densification of the tissue. This effect is much more pronounced at a concentration of 10% PVDF, and therefore, these conditions were selected for further testing.

Figure 3(a) displays the first charge-discharge profiles of the anodes with these CNT tissues. The capacity was calculated with respect to the mass of graphite loaded, and the results are summarized in Fig. 3(b). The charge curves exhibit a step around 0.8 V during the first lithiation, which does not change consequent to IPA pretreatment. Yehezkel et al.²⁴ reported on the effect of IPA on the tissue and its ability to decrease the irreversible capacity of the pristine CNT in ∼50%. However, in this work, the irreversible capacities (with and without IPA wash) were higher than those recorded earlier, probably due to unlike syntheses of the tissues.

Pretreating the CNT with a PVDF/NMP mixture has resulted in densified tissue. Nevertheless, it had no effect over the step recorded at 0.8 V or the irreversible capacity loss. Moreover, the maximum reversible capacity achieved after the first cycle is lower than for a pristine CNT current collector. Presumably, the PVDF forms a layer on top of the CNT surface which hinders the de-intercalation of Li, thus increasing the capacity loss.

CNT tissues were pretreated with strong oxidants, H₂O₂ (10%-30%) and H₂SO₄ (10%-40%), in order to remove any residual organic contaminants. HR-SEM images, displayed in Fig. 4, confirm that these pretreatments, on either concentration, have no noticeable effect on the surface morphology. EDS measurements confirm that the chemical composition, specifically the content of inorganic contaminants, did not change either.

Raman spectra of CNT tissues were recorded in order to assess the concentration of defects in the crystalline structure of the tissue. The Raman spectrum of CNT has two characteristic peaks of special interest.^30,31 The G band, at ∼1582 cm⁻¹, is attributed to stretching of the C–C bond in sp² carbon systems (graphite and its derivatives). The D band, at ∼1345 cm⁻¹, is ascribed to disorders and defects in the sp² graphene structure. The ratio between the intensities of these two bands (I_D/I_G) is proportional to the ratio between the content of those defects and the ordered, crystalline CNT. Therefore, as the tissue is cleaner, the content of organic, non-crystalline carbon is lower, and this ratio is expected to decrease.

Figure 5(a) displays exemplary spectra recorded for CNT washed in 10% H₂SO₄ or 15% H₂O₂. The ratio of I_D/I_G was calculated from those spectra and is presented in Figs. 5(b) and 5(c). As can be seen, only minor dissimilarities in the content of carbon defects are observed before and after any pretreatment.

CNT current collectors pretreated with H₂O₂ exhibited similar behavior to the pristine CNT upon the first charge-discharge cycle, as displayed in Fig. 6. By contrast, a lower irreversible capacity during the first cycle was recorded using the H₂SO₄ pretreated CNT current collector, probably due to a chemical modification following the acid wash. The step recorded at 0.8 V during the first lithiation remains unaffected with and without both pretreatments. This indicates they have no effect over the SEI formation. However, the total capacity, proportional to the amount of intercalated Li, is lower after washing with acid and closer to the capacity a Cu current collector provides. It is suggested that the intercalation of Li in the inner core of the nanotubes is diminished by this pretreatment; hence, less Li stays within the CNT after de-lithiation, and the irreversible capacity loss is diminished.

To conclude, CNT tissues were appraised as an anode current collector for the Li-ion battery, by means of morphological, chemical, and electrochemical analyses. The irreversible capacity recorded at the first cycle was diminished in two different methods. High loading mass of active material prompts the formation of the SEI on the graphite instead of on the CNT, thus decreasing the active surface area and the capacity loss originated from its formation. A decrease in the capacity loss during the first cycle has also been achieved by washing the CNT tissue with H₂SO₄ prior to anode preparation. Supposedly, this is due to the minimization of the irreversible intercalation of Li in the inner core of the nanotubes during the first lithiation. These encouraging results mark CNT tissue as a promising contender for the anode current collector in Li-ion batteries.

The authors would like to acknowledge the support of the Israeli Ministry of Energy and Water, the Planning & Budgeting Committee/Israel Council for Higher Education (CHE) and the Prime Minister Office Fuel Choice Initiative, within the framework of “Israel National Research Center for Electrochemical Propulsion” (INREP), and the Grand Technion Energy Program (GTEP).