Flexible robust binder-free carbon nanotube membranes for solid state and microcapacitor application

Carbon nanotubes (CNTs) are candidates for electrode application in energy storage and conversion systems: advanced batteries, supercapacitors and fuel cells [1–13]. The broad interest in CNTs stems from their unique thermal, electrical and mechanical properties, the diverse potential applications they offer, and the fact that the CNTs are ideal prototype to investigate quantum phenomena in one-dimensional systems. CNTs have been shown to exhibit high tensile strength (∼100 GPa) [14, 15] and elastic modulus (∼1.0 TPa) [14, 15], high electrical (∼10–30 kS cm⁻¹) [16] and thermal (∼1500 W mK⁻¹) [17–19] conductivities, stability at moderate temperatures, low density, high aspect ratio (as high as ∼10⁷) and intrinsic maximum surface area (internal and external surfaces) of graphene ∼2600 m³ g⁻¹ [20]. The monolayer atomic structure and accessibility of both external and internal pores provide additional channels to modify the CNT properties by either coating the surface with metal or metal oxide nanoparticles [21–24], non-covalent attachment of molecules or polymers to the surface [25–28], or infiltrating the inner pores with other chemicals [29, 30]. These structural modifications and the unique intrinsic properties have made CNTs promising candidate for numerous applications in the emerging nanotechnologies. Most of these applications require macroscopic structures consisting of cohesively packed trillions of the individual/bundles CNTs. Tremendous advances have been made in transforming the CNTs into bucky paper, mats, yarns and membranes. However, retaining the properties of the individual CNTs in these forms of macrostructured materials is a tremendous challenge. In the case of electrode processing, the use of mechanical compaction and solid state/liquid-phase processing techniques have become the norm of post-synthesis assembly of the CNTs [31–34]. In most cases, the properties of the macrostructured materials are abysmal compared to that of the individual CNTs. Other carbon sources including activated carbon [35–37], graphite [38, 39], graphene [35, 40] and carbon fiber [41] have been used and demonstrated as electrode materials for electrochemical double layer (ECDL) capacitors in aqueous [35–38, 41] and ionic [42–45] electrolytes, as well as in solid state ECDL capacitors [39, 40, 42–46].

Conventional electrodes for electrochemical energy storage and conversion systems usually consist of a mixture of the active electrically conducting material, and a binder to form a paste that is deposited on a metal current collector. The use of CNTs as electrode material in Li ion batteries, electrochemical capacitors and fuel cells [1–13] have been demonstrated. Several techniques have been used in fabricating the CNT electrodes. These involve the direct growth of the CNTs on Ni support [8, 9] or graphitic foils [8, 9] or the used of electrophoresis to deposit the CNT on a metal current collect [7]. In most cases, binders are used to hold the powder together [10–12]. The additional mass of the binder, the conducting material and/or the supporting substrate poses two major performance limitations: the increase in mass itself of the electrode, and the decrease in accessible active surface area. These limitations have been eliminated with the use of CNT 'Bucky-paper' electrodes formed from the filtration of surfactant-dispersed CNTs [31], or the densification of CNT forests via liquid-induced collapse techniques [32], or with the use of super acids [33, 34]. However, the 'Bucky paper' formed are normally flaky, fragile and brittle with low packing density, and swelling has become a major problem in most of these freestanding CNT-based electrodes. Where a binder is used, it decreases the overall performance of the device by increasing the mass and reducing the active accessible surface area. The strong acids tend to lead to functionalization of the CNTs with undesired species and damage to the CNT skeletal structure, degrading the intrinsic properties and quality of the CNTs. Thus, there is lack of a consistent, reproducible and viable scalable process in converting CNT powder into freestanding binder-free macrostructures.

We present a liquid phase post synthesis self-assembly (LP-PSSA) protocol that transforms the trillions of CNTs in powder form into densely packed binder-free macroscopic membranes with hierarchical pore structure that are flexible and robust with limited or no impact on the intrinsic properties of the CNTs, using halogens or halogenated high density organic liquids (Cargille Heavy Liquids). The binder-free CNT membranes have mass density greater than that of water. We correlate the mass of the CNTs loaded in the liquid to the thickness of the membrane formed after the fabrication process and observed a transition near 50 μm, from high flexibility to brittleness in the flexural properties of the membranes. The use of the binder-free CNT membranes as electrode was demonstrated for making both solid-state stacked capacitor as well as interdigitated electrode with very high power density and no degradation in performance after 10 000 cycles.

2.1. Fabrication of binder-free CNTs membrane

2.2. Characterization of the CNT membranes

2.3. Electrochemical testing

We show in figure 1 optical images of two sets of the CNT membranes (flexible membranes: CNT-FMs and brittle membranes: CNT-BMs) fabricated under the same conditions using the steps outlined in sections 2.1.1 and 2.1.2 above. The CNT-FMs in figure 1(a) are extremely flexible and robust. These membranes have thicknesses less than 45 μm. On the other hand, the CNT-BMs in figure 1(b) of thicknesses ≥55 μm are buckled and brittle, even though the processing protocols for both sets of membranes were the same. We will elaborate more on the differences later. For clarity, we will use CNT-FMs for the flexible membranes and CNT-BMs for the brittle CNT membranes in referring to figures 1(a) and (b), respectively.

Zoom In Zoom Out Reset image size

In figure 2(a), we show an SEM image of the surface morphology of the smooth and shiny surface (surface in contact with the petri dish) of CNT-FM. It reveals densely packed and randomly oriented CNTs in the in-plane of the membrane. Figure 2(b) is a filtered image of figure 2(a) into a two-dimensional black–white mode for clear distinction of the empty spaces (voids) at areas where there are no CNTs. This was done using threshold adjustment of an image analysis software, ImageJ [54]. The black regions correspond to the voids in the original image (figure 2(a)). A two-dimensional statistical distribution analysis of the area of the voids using the image analysis software, ImageJ [54] is shown in figure 2(c). The average total area of the voids from several SEM images irrespective of the thickness of the membrane (10 μm ≤ t ≤ 45 μm) was estimated to be ∼16% ± 1%.

Zoom In Zoom Out Reset image size

Similar analysis of several SEM images of the smooth side of the brittle CNT membranes gave an estimate of ∼17% ± 1% of voids. On the other hand, the average area of the voids on the dull side of several SEM images of both flexible and brittle CNT membranes was estimated to be ∼20% ± 1%. This difference might be due to the surface roughness of the top surface of the membranes. Figure 2(d) shows an SEM image of a broken edge of a CNT-FM that reveals densely packed and stacked layers in the membrane. Figure 2(e) is a cross-sectional SEM image of a piece of a CNT-FM of thickness ∼32 μm. The thickness of the membranes was measured with a non-distractive and non-contact optical profilometer, Zygo NV7300. The 3D, the 2D and the line profile analysis of a piece of CNT-FM using the optical profilometer are shown in figures 3(a)–(c), respectively.

Zoom In Zoom Out Reset image size

Shown in figure 3(d) is a plot of the variation of the mass of CNT in the processing liquid (m) versus the CNT membrane thickness (t) as determined from the optical profiler measurements for membranes processed under the same conditions, and assembled in the same processing dish (diameter ∼52 mm). There was very limited mass loss due to the processing, on average, about 1% of the loading mass. Each data point represent the average thickness of five membranes prepared using the same amount of mass per loading and under similar conditions. The error bars indicate the spread. The plot (solid dots) in figure 3(d) reveals a linear relationship between the mass loading (m) of the CNTs and the thickness (t) of the CNT membranes. The solid line is the least square fit of the data. The slope (s) from the fit is ∼22.58 g cm^–1. From the slope (s), we estimated the bulk mass density of the membranes to be ∼1.11 g cm^–3 ± 0.03 g cm^–3 using the expression m = ρV = (ρπD²/4)t. Where s = ρπD²/4, V is the volume of the membrane, D is the diameter of the membrane (∼diameter of the processing dish), t is the thickness of the membrane and ρ is the bulk mass density of the membrane. We have already demonstrated that the mass density of the membrane is greater than that of water (∼1.0 g cm^–3) [49]. In figure 3(e) we show the EDS spectra of the as-received CNTs (left plot) and a CNT membrane (right plot). The Kα and Lα peaks of Fe near 0.7 and 6.4 keV are almost non-existence in the spectrum collected from the CNT membrane. This was observed in all membranes irrespective of flexibility or brittleness. The removal of the residual contaminants, mainly metal catalyst is critical to the electrochemical performance of the CNT membranes as electrode in energy storage and conversion systems. The improvement in sample quality was also confirmed with TGA and Raman [49].

Interestingly, the membranes become brittle as the thickness increases. We observed that the membranes with thicknesses t ∼ ≤ 45 μm seem to be more flexible (see figure 1(a)). However, we also observed a gradual transition from highly flexible and robust membrane to buckling and brittleness as the thickness increases, as shown in figure 1(b). Strong buckling and brittleness occurred in membranes with thicknesses ∼≥55 μm (figures 1(b) and 3(d)). We observed that the thinner the membranes, the uniform the drying process. As the thickness increases, the drying becomes nonuniform, which could induce strain in the membranes, leading to the buckling. Additionally, in the thinner membranes, the CNTs are mostly oriented and interwoven in the in-plane direction. However, as the thickness of the membrane increases and approaches the length scales of the CNTs, more and more CNTs might be oriented in the out-of-plane direction, which might lead to the buckling and the brittleness during the drying process.

A comparison of the nitrogen adsorption (solid maker) and desorption (open markers) of the three types of CNT samples are shown in figure 4(a), CNT powder (square markers), CNT-FM (circle markers) and CNT-BM (triangle markers). The hysteresis in the three samples suggests the presence of meso- and macro-pores. The hysteresis is more pronounced in the CNT-BM sample. The BET surface area of the CNT powder, the CNT-BM and the CNT-FM are 744 m² g⁻¹, 798 m² g⁻¹ and 792 m² g⁻¹, respectively, as determined from the BET plot shown in figure 4(b) (CNT powder: squares, CNT-BM: triangles and CNT-FM: circles). The relatively small increase from the powder to the membranes might mostly be due the change of 'state', clusters in powder form to interwoven in membrane. This also suggests that the processing protocol did not have much effect on the CNTs skeletal structure, since we accessed only the external surfaces in both the powder and the membranes. The fact that there is no difference between the BET of the CNT-FM and CNT-BM suggests the consistency and repeatability of the protocol. Figures 4(c) and (d) show the deduced cumulative pore volume (top plot) and the differential pore volume (bottom plot) using the DFT adsorption analysis (CNT-FM: solid curve, CNT powder: dash dot curve and CNT-BM: dash curve). The CNT powder and CNT-FM show broad distribution of mesopores (20–50 Å) and large amount of macropores >50 Å. The pore size in the CNT-BM in the micropore region is shifted up a bit compared to that of CNT powder and CNT-FM.

Zoom In Zoom Out Reset image size

Figure 5 shows the TEM micrograph of microtomed crosslinked PVA-H₃PO₄ gel electrolyte membrane. The micrograph shows distinct segregation of phosphoric acid domains in the PVA matrix in the order of 1–2 μm. The phosphoric acid domains was well dispersed throughout the PVA matrix. In order to improve the mechanical properties of the membrane, small amount of phosphorylated silica (2.5 wt%) was added to the membrane. The synthesis of the nanocomposite is illustrated in figure 5.

Zoom In Zoom Out Reset image size

The synthesized silica nanoparticles were in the order of 10 nm and were uniformly dispersed in the PVA matrix. The ionic conductivity of the membranes measured at room temperature generally increased with increasing in the phosphoric acid content. The maximum content of phosphoric acid was limited to ∼67 wt% in unmodified PVA-H₃PO₄ as further addition significantly deteriorated the mechanical properties of the PVA making it harder to peel off the tape casted membranes. The thickness of the membrane was ∼80 μm. With the addition of silica, the phosphoric acid content increased to ∼72 wt% and the membranes could also be casted as thin as 20 μm. The maximum ionic conductivity achieved at 298 K for unmodified PVA-H₃PO₄ membrane and silica modified PVA-H₃PO₄ was 1.44 × 10⁻² S cm⁻¹ and 3.18 × 10⁻² S cm⁻¹, respectively. The temperature dependence on ionic conductivity of the membranes was fitted using Vogel–Tamman–Fulcher model as follows: $\sigma \,={\sigma }_{0}\exp \left(-\tfrac{B}{T-{T}_{0}}\right)$ where $B=\tfrac{{E}_{{\rm{A}}}}{{k}_{{\rm{b}}}},$ E_A is activation energy, k_b is Boltzmann constant, σ₀ is pre-exponential factor and T₀ is the ideal glass transition temperature as shown in figure 6.

Zoom In Zoom Out Reset image size

The insert table summarizes the values of various model parameters for both unmodified PVA-H₃PO₄ and silica modified PVA-H₃PO₄ membrane. For similar loading of phosphoric acid, addition of silica lowered T₀ while increasing the value of σ₀. The value of T₀ corresponds to the temperature at which ionic motion becomes completely hindered while increase in σ₀ suggests increased concentration of charge carriers. The value of E_a indicates that the predominant transport is related to proton hopping (Grotthus mechanism) [55].

The fabricated CNTs and PVA-H₃PO₄ membranes were then used as electrodes to construct all solid-state electrochemical capacitors. Figure 7(a) shows the constant current charge/discharge plot of capacitor made using 20 μm thick electrode and solid state PVA-H₃PO₄ membrane being cycled at 5 A g⁻¹. The specific capacitance of the electrode was ∼48 F g⁻¹.

Zoom In Zoom Out Reset image size

The solid state capacitor was cycled at various current densities and showed 92% capacitance retention even at high current densities of 20 A g⁻¹ as shown in figure 7(b). The solid state capacitor also showed excellent cyclability when cycled at high temperature upto 80 °C as shown in figure 7(c). In order to increase the specific capacitance of the CNT electrode, the CNT membranes were slightly oxidized using 0.002 M KMnO₄ solution and this resulted in an increase in specific capacitance of the electrode to almost 72 F g⁻¹, which corresponds to almost 50% increase in capacitance.

The solid state capacitors made using both as prepared CNT and oxidized CNT showed excellent cycling stability up to 10 000 cycles as shown in figure 8(a). We also saw that unlike the pristine CNT, the capacitance increased with increasing temperature for the oxidized CNT. The specific capacitance of ox-CNT almost increased to 100 F g⁻¹ at 70 °C while the capacitance of pristine CNT remained invariant with temperature as shown in figure 8(b). The temperature dependence of the ox-CNT shows induced pseudocapacitance introduced by the oxygen surface functional groups on CNT during KMnO₄ treatment. This is also consistent with the measurement of the time constant of the solid state capacitors which shows that with the oxidation of CNT, the time constant increases from 31.8 ms (pristine CNT) to 0.79 s (figure 8(c)).

Zoom In Zoom Out Reset image size

Figure 9 shows the electrochemical performance data of 3-stacked CNT solid state capacitors. Three solid state capacitors were stacked in series using grafoil as bipolar current collector. The stacked capacitor was capable of being charged and discharged from 0 to 3 V as demonstrated by the cyclic voltammogram in figure 9(a). Constant current charge/discharge with discharge time in the order of 10 s and good cyclability with almost 85% capacitance retention over 10 000 cycles at 70 °C was also demonstrated (figures 9(b) and (c)). The measured time constant of the capacitor was 0.3 s (figure 9(c)) and the leakage current of the capacitor when held at 2.5 V for 72 h was ∼100 nA cm⁻² (figure 9(d)).

Zoom In Zoom Out Reset image size

In figure 10(a) we show a laser scribed interdigitated electrodes with a gap space of ∼85 μm. The galvanostatic charge/discharge curve is shown figure 10(b) for a device with 45 μm thick electrodes with minimal IR drop. The cyclic voltammogram in figure 10(c) is nearly rectangular at 500 mV s⁻¹, and shows an areal capacitance reaching 26.9 mF cm⁻². The volumetric capacitance shown in figure 10(d) indicates stability of the device at ∼4.52 F cm⁻³ over 200 cycles. The areal capacitance exceeds that of CNT-based microcapacitors, however, our device needs fine tuning in the energy and power densities [40]. We obtained energy and power density of 0.399 mWh cm⁻³ and 16.88 mW cm⁻³, respectively, while Lin et al reported an energy and power density of 2.42 mWh cm⁻³ and 46 W cm⁻³ using a CNT forest on graphene current collectors [52].

Zoom In Zoom Out Reset image size

This paper presents a unique post synthesis protocols to self-assemble CNT powder into ultrathin flexible membranes that has several potential applications. The LP-PSSA technique present a unique protocol that eliminates the need for binders to hold the CNT powder together. The protocol can be used to assemble binder-free macro-size membranes of other carbon-based structures such as graphene, nanofibers, nanoparticles and their combinations. It involves a process of spontaneous self-assemble of the CNTs by a charge transfer mechanism via wet chemistry processing protocol. Most of the reported wet chemistry techniques end up functionalizing the CNTs or degrading the structural and the physico-chemical properties of the CNTs [33, 34]. Our approach uses charge transfer mechanism to induce the self-assembly. The charge transfer weakens the van der Waals interactions between the tubes, preventing agglomeration and facilitating the random interwoven assemble (figures 2(a), (b)). In addition to the liquid facilitating the self-assembly, it should have limited or no impact on the physico-chemical properties of the CNTs.

We report such charged transfer mechanism using liquids such as bromine to create an interwoven CNT membrane that is highly dense, flexible, binder-free and robust with bulk mass density greater than that of water [49]. Bromine is a mild etchant and has high electron affinity (electron accepter) that can easily induce charge transfer. It is well known that when bromine interacts with CNT, it transfers electrons from the carbon π states, creating a partial charge transfer between the CNT and the bromine [30]. This charge transfer mechanism weakens the van der Waals interactive force between the CNTs that allows uniform dispersion of the CNTs in the liquid bromine. This charge transfer mechanism prevents agglomeration and clustering of the CNTs, leading to homogeneously distributed CNTs in the bromine. The high aspect ratio of the CNT, in addition to the weakened van der Waal forces and the density of bromine being greater than that of the CNTs, cause the CNTs to self-assemble into a floating film on the surface of the bromine liquid. Removal of the excess bromine via evaporation and cryogenic trapping, results in the production of a high packing density, robust, flexible and binder-free CNT membrane. The excess bromine is easily removed at low temperatures ∼45 °C. Treatment at high temperature in the presence of hydrogen gas removes any residual bromine, other functional groups and improves the quality of the CNTs [49]. As can be seen in figure 3(e) (left image), there is no presence of bromine and chlorine Lα peaks near 1.5 keV and 2.6 keV, respectively. Furthermore, the Kα and Lα peaks of Fe near 0.7 keV and 6.4 keV are almost non-existence in the spectrum collected from the CNT membrane. These clearly indicate that the purification protocol outline in section 2.1.2 was effective in removing contaminants (bromine, chlorine, iron) in the CNT membranes. Halogenated high density organic liquids (Cargille) can also be used to assemble the CNTs. The high mass density, the SEM, the Zygo, and the BET data indicate a process with high consistency and repeatability. This is also consistent with the Raman and the TGA data [49].

The flexibility of the CNT membranes depends on the thickness of the membranes in relation to the length scales of the CNTs. The process reveals a transition in the flexural properties of the membranes from high flexibility to brittleness. The flexibility decreases gradually with increasing membrane thickness. All the membranes with thicknesses <45 μm seem to be flexible (figure 1(a)), whereas, as the thickness increases, the membranes become more brittle and buckling (figure 1(b)) after the processing. We suspect that as the thickness of the membrane increases, uneven stresses are induced in the membrane due to the nonuniformity in the drying of the membrane and the increase in out-of-plane oriented CNTs, which could lead to the brittleness and the buckling. The nonuniformity in the drying pronounces with increasing membrane thickness.

The CNT membranes show hierarchical pore structure (micro-, meso- and macropores) as shown in figure 5. We have used such membranes as electrodes to demonstrate symmetric, oxidized and micro (interdigitated) solid state ultra-capacitors (figures 7–10) with time constants (∼32 ms) approaching other carbon based electrolytic capacitors [49–63]. The specific capacitance of electrodes with varying thicknesses from 20 to 60 μm was invariant, and was ∼72 F g⁻¹. We, thus envision the use of the fabricated ultrathin CNT-based membrane as potential current collectors or scaffolds with high chemical stability to deposit redox active materials for both high power flexible battery and supercapacitor applications.

Our proposed self-assembly protocol and the results reveal new insights that can be applied to other nanostructures. We have shown in this investigation that it is possible to fabricate macrosize flexible binder-free CNT membranes that are easy to handle. The flexibility depends largely on the thickness of the membrane, and decreases with increasing thickness. However, properties such as the density and the surface area are independent of the nature of the membrane, flexibility or brittleness. Charge transfer engineering of the CNTs is a critical component in achieving dispersion and a densely packed membrane with bulk mass density greater than that of water. The use of mild etchant restricted the degradation of the intrinsic properties of the CNTs in the macrosize membranes. The protocol, shows very high consistency and repeatability. The CNT membranes have the potential to be used as scaffolds to deposit other materials for use in catalysis, photovoltaic, thermoelectric and many other applications, as we have demonstrated for ultrahigh power single, stack, oxidized and interdigitated solid state ultra-capacitors.

The authors would like to acknowledge the Center for Dielectric and Piezoelectrics (CDP) and NSF Nanosystems Engineering Research Center for Advanced Self-Powered Systems of Integrated Sensors and Technologies (ASSIST) for their financial support. We would also like to thank the NSF MRSEC Materials Research Facility Network (MRFN) Faculty Program at The Pennsylvania State University Materials Characterization Laboratory (MCL) for providing the financial support for characterization facilities to study the properties of the CNT membranes, and Penn State Altoona Office of Research and Sponsored Programs for their financial support of the undergraduate students engaged in this research.