Hierarchical Assemblies of Carbon Nanotubes for Ultraflexible Li‐Ion Batteries

The flexible batteries that are needed to power flexible circuits and displays remain challenging, despite considerable progress in the fabrication of such devices. Here, it is shown that flexible batteries can be fabricated using arrays of carbon nanotube microstructures, which decouple stress from the energy-storage material. It is found that this battery architecture imparts exceptional flexibility (radius ≈ 300 μm), high rate (20 A g(-1) ), and excellent cycling stability.


DOI: 10.1002/adma.201600914
architecture. Vertically aligned CNT "forests" (Figure 1 a) have been reported previously as current collectors for high performance non-fl exible batteries. CNTs are promising current collectors for batteries because of their electrochemical stability, excellent electrical and thermal conductivity, and large surface area. [ 8,[16][17][18][19][20] However, when CNT forests are bent, they readily crack (Figure 1 e). A solution for this problem is to pattern the electrode material in strips, which allows for fl exibility in bending along the direction of the strips, however this leads to the same problems when bending in the direction perpendicular to the strips (Figure 1 b). In other words, fl exibility in two bending directions requires pillar-like geometries, but even then, the ratio between the bending radius and the pillar base should remain suffi ciently high to avoid stress at the interface between the pillars and current collector (see Figure 1 c). Therefore, this work focuses on microcones as a novel geometry, combining a slender base for stress reduction and a wide crown for particle loading, as shown in Figure 1 d. A further motivation for using cones instead of pillars is that CNT pillars form unpredictable wrinkles during the capillary aggregation steps used in our battery fabrication process (see further). [ 21 ] SEM images of plain CNT forests bent to a radius of 3 mm already show cracks (Figure 1 e), while microcone electrodes bent to radii as small as 300 µm remain intact (Figure 1 f). Figure 2 depicts the electrode fabrication process, which starts by lithographically patterning catalyst particles into rings from which CNTs are grown by thermal chemical vapor deposition (CVD) (see methods and Figure 2 a). The CVD process results in the formation of microcylinders that each consists of thousands of vertically aligned CNTs (Figure 2 b). Next, these cylinders are transformed into cones using elastocapillary aggregation (Figure 2 c). [21][22][23] Shortly, this process uses capillary forces to pull the CNTs into a close packing, which results in an overall compaction of the CNT cylinders into cones (Figure 2 d and Figure S1, Supporting Information). [ 22 ] The cones are then transferred by contact printing [ 24 ] to a fl exible conductive fi lm (Figure 2 e) consisting of poly(vinydlene difl uoride) (PVDF), double wall CNTs, and phenyl-C61-butyric acid methyl ester (PCBM) in a ratio (90:5:5), which are thoroughly mixed using a planetary ball mill and cast in ≈15 µm thick fi lms, see methods for details. The fl exible fi lm is referred to as PCP ( P VDF, C NT, and P CBM) in what follows. A top SEM view (Figure 2 i) of the fi lm shows that the CNTs bundle into rafts similar to previous reports, [ 25 ] and cross-sectional SEM images ( Figure S2, Supporting Information) show a uniform spread of material through the fi lm thickness. Interestingly, our PCP fi lm (1 cm 2 ) is found to be ≈20 times lighter in weight on comparison with standard Cu-foil current collector (MTI corporation Code EQ-bccf-25 µ) ( Figure S3, Supporting Information). Figure 2 f,j shows that the CNT cones can be transferred with yields close to 100%. Further, it is important to note that most of the catalyst particles remain One of the most fascinating paradigm shifts in modern electronics is the fabrication of soft and fl exible electronic devices. This development is fueled by a continuous search for more compact and intuitive consumer electronics, [ 1 ] medical implants, [ 2 ] and the emergence of the "Internet of Things." [ 3 ] While considerable progress has been made in the fabrication of fl exible and stretchable circuits, [ 4 ] radio-frequency identifi cation (RFID) tags, [ 5 ] and displays, [ 6 ] the fl exible batteries needed to power these devices remain challenging. [ 7 ] Existing fl exible batteries are often too heavy, bulky, and rigid, and require a radical redesign of the battery architecture to address these issues. [ 8 ] Here we demonstrate that extremely fl exible batteries can be achieved by designing carbon nanotube (CNT) microstructures, which decouple the stress induced during bending in the collector electrode from stress in the energy-storage material (Fe 2 O 3 anodes and LNCO cathodes in this work). We found that this battery architecture not only imparts excellent fl exibility (bending radius ≈ 300 µm), but also high rate (20 A g −1 ), cycling stability (over 500 cycles at 1 C with capacity retention over 70%).
The design of highly fl exible batteries requires judicious engineering of the electrodes to mitigate stress concentration and crack formation. Recent, progress toward fl exible batteries includes the use of polymer substrates, [ 9 ] composite membranes, [ 10 ] CNT yarns, [ 11 ] paper-based electrodes, [ 12 ] graphene foams, etc. [ 8,13 ] However, many of these designs suffer from fast-capacity decay, [ 14,15 ] limited fl exibility, poor thermal management, and large weight. [ 8 ] Here, we propose a design where stress in the electrode is localized in the current collector and decoupled from the electroactive region, which remains unstressed during bending. To achieve this, we populate a fl exible current collector with microscale 3D structures, which contain the electrochemical active material. In order to remain unstressed as the electrode bends, the base of each active microstructure must be small, and therefore, similar to trees, our microstructures have slender trunk planted in the fl exible current collector and a wider crown loaded with electrochemically active nanoparticles. Figure 1 further illustrates our rationale behind the design of the electrode attached to the silicon substrate after the CNT transfer printing process, and therefore the Si-substrate can be recycled to synthesize new CNT cones. With our current process the catalyst can be regrown about four times before losing its activity. [ 26 ] Finally, the electrodes are decorated with Fe 2 O 3 nanocrystals (≈10 nm diameter, Figure 2 g, and Figure S5, Supporting Information) by drop-casting a carefully weighed hexane-Fe 2 O 3 suspension. The iron oxide nanocrystals are synthesized by a colloidal method, which involves decomposing iron pentacarbonyl in octadecene in the presence of oleylamine (see methods and Supporting Information). [ 27 ] Iron oxide is chosen as the anode material because of its high theoretical capacity (≈1000 mA h g −1 ), low cost, environmental friendliness, and abundancy. [ 28,29 ] In addition, the conversion reactions that take place between Fe 2 O 3 and Li ions during charge/discharge cycles require good electrical contact, which is provided by our unique CNT microcone architecture. SEM images of transferred cones on PCP fi lm, loaded with Fe 2 O 3 nanoparticles (≈0.5-1 mg) are shown in Figure 2 k,l and Figure S4 of the Supporting Information.
The electrodes are packaged, using both standard coin-cells (2032) with Li metal (see methods) and fl exible laminated packages using cathodes made using the same CNT cones but decorated using commercial, lithium nickel cobalt oxide powder (LNCO) (Sigma-Aldrich 760986) because Li metal has limited fl exibility. [ 8 ] In a fi rst series of tests we use the coin cells to compare our hierarchical microcones decorated with Fe 2 O 3 to conventional fl exible electrodes fabricated by mixing the same Fe 2 O 3 nanoparticles with the CNT based PCP fi lm. Figure 3 a shows that the charge/discharge curves of both electrodes have a sloping plateau at ≈1.8 V corresponding to the fi rst-step lithiation of Fe 2 O 3 . This becomes more defi ned at ≈1.2 V with the onset of conversion reaction and at voltages below 1.0 V, the sharp increase of current can be attributed to a full conversion coupled with the formation of a solid electrolyte interface (SEI). [ 30 ] These processes can also be seen in the cyclic voltammogram (Figure 3 a inset) with a gradual lithiation/delithtaion of Fe 2 O 3 at 1.8, 1.2, 0.5, and 2.0 V, in accordance with typical reports on conversion of Fe 2 O 3 (i.e., Fe 2 O 3 + 6Li → 2Fe + 3Li 2 O). [ 29,31,32 ] The peak at 0.7-1.0 V corresponding to Li + ion insertion into Fe 2 O 3 is more pronounced in the CNT-cone electrodes, suggesting a facile lithiation reaction.
The initial capacity of the CNT cones is ≈1000 mA h g −1 at 0.5 C while the conventional electrodes yielded only ≈800 mA h g −1 (Figure 3 a discharge curves), but in the subsequent cycles the conventional electrode loses more than half of the initial capacity (Figure 3 b), while the CNT cone electrodes remarkably retain more than 70% of their initial capacity. At high current densities (rates), CNT cone electrodes clearly outperform the capacity and retention of the conventional fl exible electrodes (Figure 3 b). For instance, the capacity of conventional electrodes quickly fades and reaches values close to 0 mA h g −1 at a rate of 7 C (Figure 3 b), which indicates poor electrochemical activities while the microcone electrodes maintain more than 300 mA h g −1 under the same conditions. Such drop in performance of conventional electrodes where the metal oxide is blended in a polymer binder was also reported previously. [ 33 ] In comparison, the CNT cone electrodes still yielded appreciable capacity at 10, 15, and 20 C (Figure 3 c), and despite these harsh Adv. Mater. 2016, 28, 6705-6710 www.advmat.de www.MaterialsViews.com high rate tests, the electrodes still exhibited sustained capacities 650-800 mA h g −1 when the rate was reduced to 0.5 C. We initially observed a difference in charge and discharge capacity after reverting from high to low rate, which might be due to the binder free nature of our electrode architecture and the thick deposition of active particles built up of multiple layers. The cone architecture therefore not only imparts excellent fl exibility, but also enhances battery performance substantially. Because the cones bring the active material outside of the binder, the electrochemical reactions take place at the surface of the CNT cones. This protects the collector electrode from the deleterious side reactions between active materials (Fe, Li 2 O, and Fe 2 O 3 ) and binder molecules, which can lead to large internal resistance. Furthermore, the developed CNT cones are able to accommodate the drastic volume change [ 34 ] taking place during conversion reactions, and ensure good contact to the active material resulting in the observed higher charge retention, also compared to other recently published CNTs-metal oxide electrodes. [ 33,35,36 ] Finally, the cone electrodes were cycled 500 times at 1 C with no appreciable change in the capacity as shown in Figure 3 d, and, in Table S1 in the Supporting Information, we further compare the performance of our battery to those based on CNT forests previously reported. Finally, we have constructed fl exible full cells with both CNT cone anodes and cathodes, as depicted in the schematics of Figure 4 a. The former is coated with Fe 2 O 3 as discussed above, and the cathode with commercial LNCO powder (see methods). Figure 4 b shows the charge/discharge curves from the full cell. The full cells were folded multiple times and then tested in fl at state, they operate around 3.1 V with a sloping plateau between 4.0-2.0 V and the capacity was stabilized after the second cycle (for LNCO), and delivers a reversible capacity of ≈120 mA h g −1 at 2 C (Figure 4 b). Figure S6 of the Supporting Information shows further cycling data of the battery while it is being bent and released for 15 cycles. These batteries were connected to a white light emitting diode (LED, 3 V) and fl exed to radii of about 3 mm while powering the LED as illustrated in Figure 4 c. The bending radius is currently limited by the Adv. Mater. 2016, 28, 6705-6710 www.advmat.de www.MaterialsViews.com packaging rather than the electrodes themselves (more information in Supporting Information). Finally, complex electrode architectures are often not able to survive the harsh battery cycling conditions. Therefore, we opened our fl exible cells after 1000 cycles, which corresponds to the lifetime of the battery to image the CNT cones. Figure 4 e reveals that the overall cone morphology is well maintained, similar to as fabricated CNT cones shown in Figure 4 d. We believe this is because the cones are well anchored in the collector electrode, and because capillary aggregated CNT structures are substantially more resilient than as-grown forests. [ 21 ] In conclusion, this paper presents a new electrode architecture, which consists of hierarchical cone shaped CNT structures with a slender trunk embedded in a fl exible light weight collector electrode and a wide crown that is loaded with nanocrystals for energy storage. This architecture allows the electrodes to be folded to very small radii without inducing stress in the active material network. These unique electrodes not only alleviate stress, but also bring the active particles outside of the binder, which dramatically enhances the performance of the conversion reactions. This results in batteries that are extremely fl exible (300 µm radii) and at the same time offers excellent rates (as high as 20 A g −1 ), and cyclability (over 500 cycles at 1 C with capacity retention >70%).

Experimental Section
Lithography : Clean (100) silicon wafers were fi rst dehydrated at 200 °C, and coated with an adhesion promoter (Ti-prime). Then AZ 4533 photoresist was spin coated (3000 rpm, 30 s) and prebaked at 115 °C for 2 min on a hot plate, and exposed. The patterns were developed in diluted AZ 351B (1:4 DI water, 3 min), followed by rinsing in deionized (DI) water and blow-drying.
CNT Growth : E-beam evaporation was used to deposit a 10 nm Al 2 O 3 and 1 nm Fe catalyst fi lm. The Si-wafers were then diced and the catalyst was patterned by lift-off. Next, CNT forests were grown in cylindrical structures by thermal CVD in a horizontal tube furnace at atmospheric pressure, with fl ows of 100/400/100 sccm C 2 H 4 /H 2 /He, at 800 °C. The CNTs were rapidly cooled in the growth atmosphere before purging the CVD chamber with helium.
Densifi cation, PCP Film Fabrication, and Transfer Printing : The cylindrical CNT microstructures were converted into elongated cone-like structures by a capillary forming process, reported previously. [ 22 ] These CNT microcones were transferred onto the fl exible PCP fi lm, which was fabricated as follows: 15 mg of methanofullerene phenyl C61 butyric acid methyl ester (PCBM) was dissolved in 3 mL dimethylformamide (DMF) and ultrasonicated for 1 h. 15 mg double-wall carbon nanotubes (DWCNTs) (Nanocyl NC2000) were added to the PCBM solution and ultrasonicated for 30 min. A solution of poly(vinylidene fl uoride) (PVDF) (300 mg in 3 mL DMF) was added as binder to the CNT-PCBM solution and stirred. Finally, the PVDF-CNT-PCBM suspension was ball milled (planetary) in a grinding jar (25 mL) with a single metal ball for 2 h with an interval of 30 s after every 2 min. The ball-milled PCP suspension was drop-cast over a clean soda-lime glass slide. The temperature of the glass substrate was increased to 60 °C to dry the fi lm and was then raised to 175 °C to soften the PVDF. The densifi ed CNT cones were then transferred on the PCP fi lm by microcontact printing. [ 24 ] Nanocrystal Synthesis and Coating : Iron oxide (Fe 2 O 3 ) nanocrystals were synthesized by a colloidal method, which involved decomposing iron pentacarbonyl (Fe(CO) 5 ) (≈1-2 mL) in octadecene (≈20 mL) in the presence of an oleylamine and oleic acid mixture (≈1.5 mL) as reported previously. [ 27 ] More synthesis details are provided in the Supporting Information, along with an XRD scan ( Figure S5, Supporting Information). Purifi ed iron oxide nanoparticles, suspended in hexane, were drop-cast onto the CNT microstructures at 40 °C. The substrate was left in air for a few minutes to completely evaporate the solvent. The PCP fi lm with CNT microstructures was then peeled-off from the glass substrate. The fl exible electrode was tested in half cells with 2032 coin Adv. Mater. 2016, 28, 6705-6710 www.advmat.de www.MaterialsViews.com www.advmat.de www.MaterialsViews.com type casings with pure Li metal as the reference and counter electrodes, and was separated by a layer of polypropylene (PP). 1 M LiPF 6 served as the electrolyte. The battery fabrication and equilibration of cells were carried out in an argon (Ar) fi lled glovebox at room temperature. The full cells were fabricated by drop-casting Fe 2 O 3 nanoparticles and commercial LNCO powder (from Aldrich 760986, <0.5 µm) on CNT cone electrodes. These were separated by a polypropylene layer or Whatman borosilicate paper soaked with 1 M LiPF 6 as electrolyte, and the pack was laminated in an Ar fi lled glovebox.
Conventional Electrode Fabrication : For comparison, conventional fl exible electrodes were fabricated by adding the required amount of Fe 2 O 3 nanoparticles (in hexane or chloroform) to 1-3 mL of presynthesised and ball-milled PCP suspension (DMF). The composite suspension was sonicated and then stirred for 1 h each before drop-casting into fi lm and packaging the battery in the same way as described above.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.