Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/nanoenergy

Nano Energy (2015) 15, 33–42

http://dx.doi.org/12211-2855/& 2015 E

nCorresponding auProcesses, Seoul Nat

E-mail address: j1These authors co

RAPID COMMUNICATION

Omnidirectionally stretchable, highperformance supercapacitors basedon a graphene–carbon-nanotube layeredstructureInho Nama,b,1, Seongjun Baea,b,1, Soomin Parka,b, YoungGeun Yooa,b, Jong Min Leeb, Jeong Woo Hanc, Jongheop Yia,b,n

aWorld Class University Program of Chemical Convergence for Energy & Environment, Institute ofChemical Processes, Seoul National University, Seoul 151-742, Republic of KoreabSchool of Chemical and Biological Engineering, Seoul National University, Seoul 151-742, Republic ofKoreacDepartment of Chemical Engineering, University of Seoul, Seoul 130-743, Republic of Korea

Received 7 January 2015; received in revised form 18 March 2015; accepted 2 April 2015Available online 14 April 2015

KEYWORDSActin–myosin interac-tion;Carbon nanotube;Graphene;Homogeneous inter-face stress;Stretchable electro-nics;Stretchablesupercapacitor.

0.1016/j.nanoen.2lsevier Ltd. All rig

thor at: World Clional University, Syi@snu.ac.kr (J. Yntributed equally

AbstractThe development of stretchable energy storage systems for fully power-independent and stretchabledevices for the next generation is increasing. Here, we report on a graphene–carbon-nanotube-layeredstructure for use as a stretchable electrode and its application in all-solid-state stretchablesupercapacitors and various electronics. In this system, graphene serves as a floating track and carbonnanotubes convert external stress into the stretching motion of the electrode. The structure providesomnidirectional deformation without inhomogeneous interface stress and slip stress between activesites and the stretching passive components. The suggested system offers significant improvement overexisting methodologies for fabricating stretchable energy storage systems and electronics in terms ofdensity of capacitance, negligible passive volume, biaxial and twisted deformation, and durability. Theintegration of stretchable electrodes in various substrates and their application as all-solid-state,stretchable supercapacitors are demonstrated, and a high value of capacitance in the deformed stateof 329 F g�1 was achieved (based on mass of the graphene). The physical characteristics of the systemare also revealed by first-principle calculations and three-dimensional finite-element methods.& 2015 Elsevier Ltd. All rights reserved.

015.04.001hts reserved.

ass University Program of Chemical Convergence for Energy & Environment, Institute of Chemicaleoul 151-742, Republic of Korea. Tel.: +82 880 7438.i).to the work.

I. Nam et al.34

Introduction and CNT cluster are bonded to form a two-dimensional

Ingenious ideas for fabricating stretchable wavy structuresor materials for use in stretchable electronics have beenproposed [1]. However, strategies that have been proposedare directed toward only phenomenological characteristicsof soft materials with obvious limits, such as the signifi-cantly large passive volume of the stretched structure andthe low and irreversible performance of stretchable materi-als [2,3]. Conventional technologies for stretchable electro-nics are composed of a stretchable bridge and a rigidoperating site. Stretchable bridges have successfully con-ferred deformability to these electronics. However, signifi-cant interfacial resistance between the stretchable bridgeand the rigid operating site is common, which fundamen-tally decreases operational stability and durability.Recently, researchers have begun to investigate stretchablesupercapacitors in their primary stages [4–8]. Energy storageis of critical importance because it is a limiting factor inachieving complete and independent stretchable electro-nics for the next generation. Moreover, supercapacitor isone of excellent alternative energy storage systems becauseof its high power density and device reliability. However,the system also does not circumvent above general limits ofthe conventional stretching strategies.

To overcome the limitations of stretchable energy storagesystems and electronics, we presented first-principle calcula-tions and experimental proof for graphene–carbon-nanotube(CNT)-layered electrodes, a system inspired by the underl-ying principle of the motion of human muscles. Graphene isone of the best candidate materials in electronic systems, e.g.,printed circuit boards, thin-film transistors, light-emitting dev-ices, solar cells, and various sensors [9]. Especially, the low-dimensional material is also considered to be the best candi-dates currently available for use in supercapacitor electrodesbecause of their high interface charge separation [10–12]. Tofabricate a stretchable graphene structure, several studies havemade use of conventional techniques: fabrication of a graphenewavy structure and intensification of its stretchable properties.However, the graphene is brittle, and its advantages, i.e., highelectric, mechanical, and thermal properties, diminish abruptlyand irreversibly in the deformed state [3,13]. In conclusion, theobjective of this study was to investigate the use of the rigidand brittle form of graphene as an operating site for stretch-able electronics in an attempt to eliminate the interfacial res-istance.

A typical cell in living bodies that undergo semi-permanent and reversible stretching is the sarcomere inmuscle tissue [14]. Sarcomeres include myosin moleculesthat function as molecular motors and actin filaments thatact as tracks that direct the motion of the motor assembly[14,15]. In the same manner, undeformable graphene canact as a track, and deformation can be accomplished using astretching component as a motor riding on the graphenetrack. It is reasonable to use a CNT cluster as the motorbecause of their significant stretchability and stability[16,17]. CNT cluster can be attached to the graphenesurface with van der Waals interactions caused by graphiticAB stacking [18]. In this system, the graphene serves as afloating track and the CNT cluster converts external stressinto the stretching motion of the electrode. The graphene

homogeneous interface, which can solve the problem ofinhomogeneous interface stress. The graphene–CNT struc-ture also has negligible slip stress and no passive surface forstretching utilization. Thus, this structure shows multifunc-tionality that was previously unachievable, i.e., omnidirec-tional stretching and energy storage via the formation of alarge surface electrochemical double-layer.

Experiments

First-principles calculations and three-dimensionalfinite-element method modeling

First-principle calculations were carried out on the basis ofdensity functional theory (DFT) using a generalized gradientapproximation (GGA) within the Perdew–Burke–Ernzerhof(PBE) functional [19]. We used the projector-augmentedwave (PAW) method as implemented in the Vienna ab-initiosimulation package (VASP) [20]. The van der Waals interac-tions are described via the DFT-D2 Method of Grimme [21]. Aplane-wave basis set with a kinetic-energy cutoff of 600 eVand a 2� 2� 10 Monkhorst–Pack k-point mesh was used. TheCNTwas built into a single-walled armchair nanotube with a(10,10) index (diameter, 13.7 nm), as shown in Figure 1a.The structures are located in the orthorhombic unit cellwith a vacuum gap. Lattice constants are a=34.08 Å, b=45.00 Å, and c=2.46 Å (see the Supplementary materials,Figure. S1). Before calculating the binding energy of thegraphene–CNT interface, we calculated the variation of theenergy along the z-axis by modifying the distance betweenthe graphene and CNT from 5 to 2 Å in steps of 0.2 Å, andthe binding energy was maximized at 3.1 Å with structurerelaxation (Figure S1b). The structure relaxation wellrepresented the covering state of graphene on the CNT.Calculated energy values were summarized in Table S1. Wealso developed three-dimensional (3D)-finite element met-hod (FEM) models for numerical structure analysis [22]. Allelements of the models were designed by means of a20-node (Serendipity) hexahedral brick. Near the edges ofthe model, we used an element size under 2 μm for highaccuracy.

Fabrication of graphene–CNT stretchable electrodes

To fabricate the graphene–CNT-layered structure, variouselastic supports [rubber, latex, poly(vinyl alcohol) (PVA),etc.] were initially stretched in two dimensions and fixed.Then 10 layers of a 0.5 g/L solution of CNT (ILJIN Nanotech.)ink dissolved in 200 μL of ethanol was spread onto theprestrained substrate. The CNT was pretreated in a 70%HNO3 solution at 110 1C for 8 h, which resulted in thesurface of the CNT being slightly hydrophilic. The HNO3

treated CNT contains some O-functional groups (8.15 wt%).However, the presence of these did not affect the majorinterfacial characteristics of the system. Following thisprocess, three layers of a 0.5 g/L graphene ink solution in200 μL of ethanol was spread onto the CNT layer. Thegraphene was synthesized from graphene oxide in a micro-wave oven for 2 min, as suggested in a previous study [23].

Figure 1 Atomic-scale properties of graphene–CNT and actin–myosin interactions. (a) SEM image of graphene–CNT interface.(b) Models for the progressive movement of myosin along actin in a living cell and of CNTs along graphene in an inorganic system.(c) The minimum energy path for the movement of myosin (or CNT) on actin (or graphene) for a living cell and inorganic system.(d) Differential charge densities ρdiff. =ρtotal�(ρgraphene+ρCNT) of graphene–CNT interface with AB and AA stacking betweenhexagonal carbon chains. Positive (yellow) and negative (blue) values are plotted on the same isosurfaces.

35Omnidirectionally stretchable, high performance supercapacitors

The pre-strained substrate was then released, resulting inthe formation of a wrinkled graphene–CNT layer.

Fabrication of electric surgery glove and contactlens

To show the electric circuit on surgery gloves and contactlenses, commercial light-emitting diode (LED) cells [HSMC-C150 (3.2 mm� 1.6 mm), AVAGO Tech.] were mounted andmanually bonded to the Ag paste anchor using a conven-tional soldering method on commercial supports. Theapplied voltage to the LEDs was a constant 2 V, suppliedan external power supply. The maximum forward voltageand power dissipation of the LEDs were 2.4 V and 60 mW,respectively. A cyclic stretching test using surgery gloves

demonstrated that the inserted LED circuit endured 100cycles with no defects.

Fabrication of stretchable supercapacitor

A PVA/H3PO4 gel electrolyte was prepared by mixing 5 g ofPVA powder, 4 g of H3PO4, and 40 mL of water together[22,24]. The mixture was heated to 358 K under vigorousstirring until the solution became clear. The gel electrolytewas poured onto a petri dish to make the PVA/H3PO4 film.After drying, the film was two-dimensionally stretchedand fixed. A 200 μL aliquot of the graphene ink (0.5 g/L)was spread three times and 200 μL of CNT ink (0.5 g/L)was spread 10 times onto the film. The same amount of inkwas spread onto the opposite side. The stretched film was

I. Nam et al.36

released, resulting in the formation of a stretchable super-capacitor system.

Morphology and electrochemical characterization

Scanning electron microscope (SEM) images of the gra-phene–CNT structure were obtained using a field-emissionSEM (AURIGA, Carl Zeiss). Cyclic voltammetry (CV) andcharge/discharge (C/D) curves were carried out using acomputer-controlled potentiostat (ZIVE SP2, ZIVE LAB) at anoperation voltage range of 0–0.8 V. All CV and C/D curves inthis study are representative and each datum was collectedat the 10th cycle of CV or C/D analyses.

Results and discussion

Figure 1 schematically summarizes the interfacial interac-tions of the graphene–CNT-layered structure. The interac-tion and moving motion of the graphene–CNT electrodes canbe demonstrated by comparing them to the actin–myosininteraction in muscles. In living bodies, a single motor in thesarcomere (myosin) moves along the track (actin) andoperates via cycling states with different affinities[14,15]. The cycling motion accompanied by a change inaffinity also occurs in the graphene–CNT-layered structure[ΔE, 0.08 eV U C�1 (Figure 1c)]. During biological musclemovement, the head of the myosin forms a chemical bondwith an actin molecule on the thin filament (crossbridge). Inthe graphene–CNT electrode system developed here, thecrossbridge is formed by van der Waals interactions betweenclosed chains of six carbon atoms in the CNT filaments andgraphene sheets with the graphitic AB stacking [18]. For thecalculation of the binding and drifting properties on thegraphene–CNT interface, the graphene–CNT structure wassimulated with DFT-optimized lattice constants [25,26]. In agraphene structure, the distance between neighboringcarbon atoms is 1.423 Å, which agrees with the experimen-tal value of �1.42 Å [27]. The binding energy, Ebinding at theinterface of the graphene and CNT was defined by

Ebinding ¼ ðEgrapheneþECNTÞ–Etotal ð1Þwhere Etotal is the total energy of the system containing theadsorbed CNT and graphene, Egraphene is the total energy forthe optimized bare graphene surface, and ECNT is the totalenergy for the CNT in the vacuum. With this definition, thepositive binding energies correspond to energeticallyfavored states (Figure 1c). In a sarcomere, as soon as thecrossbridge is formed, inorganic phosphate (Pi), producedby the hydrolysis of adenosine triphosphate (ATP) in themyosin, is released with the concomitant motion of thelever arms (power stroke) [14,15]. In the graphene–CNT-layered system, the shear stress from external pulling ortwisting between the graphene and the CNT replaces thechemical energy from hydrolysis of ATP. The binding energyof graphene–CNT AB stacking is 0.49 eV U C�1. When theCNT carbon atoms slide on the graphene (hopping mechan-ism), the activation energy from AB stacking to metastableAA stacking is only 0.08 eV U C�1 (Figure 1d). This value issix times lower than the binding energy of the graphene–CNT network, which induces the CNT motors to move alongthe graphene layers rather than detach. Especially, the

interface stress (slip boundary) between the CNTcluster andthe graphene layer has a negligible influence on the overallstrain of the inorganic muscle because the energy needed tomove the CNT along the graphene track is only 0.08 eV UC�1, which is much smaller than the external stress. Thegraphene–CNT-layered structure is homogeneously locatedon the two-dimensional surface, which causes uniformstrain of the electrode surface, enhancing the durability.

The differential charge density, ρdiff., of the graphene–CNT interface is defined by

ρdiff: ¼ ρtotal–ðρgrapheneþρCNTÞ ð2Þwhere ρtotal is the charge density of the graphene–CNTinterface, ρgraphene is the charge density for the baregraphene surface, and ρCNT is the charge density for theCNT. The charge accumulations (yellow) and charge reduc-tions (blue) around the CNT and graphene are explicitlyshown in Figure 1d. The two different charge densities sharethe same feature: charge accumulations and reductions arevertically superposed between the CNT and the graphene,which suggests an electrostatic interaction at the graphene–CNT interface. In this system, the interaction energydepends on the van der Waals interaction between thegraphene and the CNT. In the case of AB stacking, electronichybridization was found at the graphene–CNT interface. Thisinduced a charge redistribution, which largely enhanced thestrength of the electrostatic interactions. In AA stacking,however, no appreciable electronic hybridization betweenthe graphene and the CNTwas detected, indicating that thestate was unstable and that the graphene–CNT interface wasspontaneously reconstructed in the form of AB stacking.Also, the CNT surface atoms naturally moved from holes inthe surrounding carbon atoms of graphene to other holesinside the rings of graphene. As a result, the CNT movedomnidirectionally along the two-dimensional grapheneplane by the continuous docking (AB)/undocking (AA) ofthe CNT edges onto the graphene flake (power stroke,Figure 1d). In biological systems, after the release ofadenosine diphosphate (ADP), the myosin re-combines withATP. The combination induces the collapse of the coupledmyosin and actin (Figure 1b). Simultaneously with thehydrolysis of ATP, the myosin and actin are randomlyreconstructed and the crossbridge is shifted by approxi-mately 11 nm, the length of neighboring binding sites ofmyosin. The result is analogous to the re-docking of the CNTand graphene [15]. In the case of the graphene–CNT powerstroke, the distance of the shift in the CNT chains wasapproximately 2.46 Å, which represents the length betweenthe neighboring holes of the six-membered rings of gra-phene (Figure S1).

To experimentally prove this concept, we obtainedoptical and SEM images and 3D-FEM simulation of thegraphene–CNT stretchable electrode at various deformationstates. Random myosin movement macroscopically appearsin the compression and spontaneous release of muscles inthe human body (Figure 2a). In contrast, the movement ofthe CNT on the graphene track yielded stretching andspontaneous release of the graphene–CNT stretchable elec-trode. A schematic diagram of the flow process for thedrifting motion of graphene is shown as SupplementaryFigure S2. The modulus and tensile strength of a singleCNT was determined to be �910 GPa and �0.15 GPa,

Figure 2 Microscale structure of graphene–CNT stretchable electrode. (a) Compressing and releasing of muscle in a human bodythrough actin–myosin interaction. (b) Optical images and equivalent stress simulation (finite element model (FEM) calculation) ofgraphene–CNTelectrodes at uniaxial, biaxial, and twisted deformation. (c) SEM images of graphene–CNT-layered structure at biaxialstretching (80%) and releasing. (d) High-resolution SEM images of individual crumpled structures in graphene–CNT networks at thepoint of stretching and releasing. The graphene layers are undeformable and separated from each other. The CNT cluster under thegraphene flakes are randomly bending and stretching.

37Omnidirectionally stretchable, high performance supercapacitors

respectively, indicating the severe stiffness of a singleanisotropic CNT [28]. However, the effective modulus ofthe CNT cluster was reduced by several orders of magnitudebecause of the non-coplaner and omnidirectional wavinessof the CNTs in a cluster. The morphological characteristics ofa CNT cluster is a factor that provides stretchability to theelectrode system. Previous research indicates Young's mod-ulus of 565 KPa and a tensile strength of 170 KPa for a CNTcluster, indicating a high elasticity with a �30% yield strain[16]. In addition, the three-dimensional random networks inthe CNTcluster enable us to assume nearly isotropic proper-ties [29]. Figure 2b shows optical images and the corre-sponding 3D-FEM modeling of graphene–CNT structuresformed after 80% uniaxial, biaxial, and twisting deforma-tion. In the graphene–CNT stretchable electrode, the gra-phene is undeformable and undergoes drifting motion. Theslip boundary stress between the CNT cluster and thegraphene layer converges to zero based on DFT calculationresults, as stated above. Therefore the FEM calculationcan be based on the CNT cluster. The value of elasticityshows a relatively low compliance value as compared to our

experiments (yield strain, �75%). This difference is causedby the synergetic effect of the inherent nanostructure ofthe CNT cluster and the crumpled macrostructure of thegraphene–CNT electrode, which enhances the omnidirec-tional stretchability (Figure 2b). The method for fabricatingthe crumpled structure is shown in the methodology asso-ciated in Supplementary Figure S3.

To calibrate the modulus of the graphene–CNT stretchableelectrode with combination effects, we used the experi-mental value for yield strain, �75%, and the fixed tensilestrength of the CNT cluster, 170 KPa. From Hooke's law, thecalibrated Young's modulus of the system was calculated tobe 227 KPa. The graphene–CNT-layered structure had adepth of 2 μm, and we designed the mesh of the structureas a low-density orthorhombic solid. For the FEM calcula-tion, the stress, σ, and strain, ε, of the materials arederived as

σi ¼ Dij � εi ð3Þwhere Dij is the elasticity matrix, which is determined byfollowing equation for isotropic materials:

I. Nam et al.38

Dij ¼E 1�νð Þ

1þνð Þ 1�2νð Þ

1 γ γ

γ 1 γ

γ γ 1

0 0 0

0 0 0

0 0 0

0 0 0

0 0 0

0 0 0

β 0 0

0 β 0

0 0 β

26666666664

37777777775; σi ¼

σxx

σyy

σzz

σxy

σyz

σzx

26666666664

37777777775; εi ¼

εxx

εyy

εzz

εxy

εyz

εzx

26666666664

37777777775

ð4Þ

where γ=ν/(1–ν), β=(1–2ν)/2(1–ν), E is Young's modulus, andν is Poisson's ratio. Poisson's ratio of the CNT cluster is0.4 from previous research [29]. From an equivalent stresscalculation using 3D-FEM, all structures (biaxial and uniaxialdeformation with �80% strain and twisted deformation witha 3601 angle) maintained their elasticity, excluding a smallportion of the edges of the structure (Figure 2b). The resultsshow that the graphene–CNT electrode design is capable ofsupporting stable and robust operations during large-scaleand mixed modes of deformation similar to the bending ofbiological muscle. To develop insight into the structuralfailure of the model, we used the Huber–von Mises failurecriteria as [22]

Safety factor=Y/ σ,

σ ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi12

σxx�σyy� �2þ σyy�σzz

� �2þ σzz�σxxð Þ2h i

þ3σxy2þ3σyz2þ3σzx2r

ð5Þwhere σ is the equivalent stress and Y is the yield strength.A safety factor less than 1 indicates that the model isunstable. As shown in Supplementary Figure S4, for the mostpart each calculated model maintained a safety factor ofmore than 1. Some instability occurred at the edges, whichcaused a slight decrease of conductivity. However, thecalculation results are in good agreement with the opticalimages (Figure 2b).

The deformation of the graphene–CNT connections inv-olves the following steps. First, the system contains non-coplanar wrinkles similar to crumpled paper, and an exter-nal force causes the wrinkled structure to physically unfold(Figure 2c and S3) [6,7]. With the physical unfolding,the serpentine CNT cluster is subsequently extended alongthe graphene track with a negligible interface stress(Figure 2d). To provide clear evidence of sliding grapheneon the CNT stretchable substrate, we conducted in-situmicroscopy analyses of the composite structure (Supportingmovie). At the released state, graphene sheets are obser-vable on the surface of the composite. With appliedstretching, the graphene sheets become cracked becauseof their rigidity. In such a situation, the conductivity of thecomposite is maintained (Figure 3a), which means that thesubstrate composed of a CNT cluster provides supportbetween graphene sheets. This result indicates not onlynegligible shear stress between the graphene and stretch-able CNTs but also tenable conductivity enabled by the CNTsubstrate. Microscopically, each crossbridge in the interfacebetween graphene and a CNT layer is randomly broken withstretching of the CNT clusters. The crossbridges are con-tinuously reconstructed during the movement of the CNT,regardless of direction, because graphene flakes offer thesystem a repetitive two-dimensional (2D) track. From amacroscopic viewpoint, the reconstruction process shows

that undeformable graphene flakes recede by the stretchingof CNT filaments (Figure S2). To estimate the deformation ofgraphene under tension, we applied a series of incrementaltensile strains on a unit cell of bare graphene using thesimulation based on DFT calculations. Uniaxial tension wasapplied in the x direction and the Poisson contraction wasassumed to be zero under the tension for computationalconvenience (see the Supplementary materials, Figure S5a).Figure S5b shows the calculated stress–strain curve ofgraphene. Strain, ε, is defined by

ε¼ L=L0 ð6Þwhere L is the length of the unit cell at the x-axis and L0 isthe value at zero stress optimized by DFT calculation. Stress,σ, is the difference in internal energies from the value of thegraphene sheet at zero stress. Up to 20% strain, graphene hasan isotropic in-plane elastic response, and the lattice sym-metry is broken when this value is exceeded. The findings arein good agreement with the experimental results of theYoung’s modulus of graphene sheets, ~1.1 TPa [30,31].However, at 2% strain, the stress (0.23 eV U.C.�1) alreadyexceeds the activation energy between AB stacking and AAstacking of the graphene–CNT interface (0.08 eV U.C.�1).This fact, therefore, verifies that external stress causes theCNT to follow the undeformable graphene track rather thanthe deformation of the graphene sheet.

Supplementary material related to this article can befound online at http://dx.doi.org/10.1016/j.nanoen.2015.04.001.

In the experiment, each robust graphene area is indi-cated by the bright regions in the SEM images of Figure 2c(Pt-coated area). The size of the graphene flake is approxi-mately 100 μm2 during the stretching and releasing, which isin good agreement with the Supporting Movie of thegraphene–CNTelectrode at various deformations. The movieshows floating and undeformed graphene flakes that aresimilar to graphene with cracks at irreversible stretchingdeformation in previous research [13]. In general, thecracks cause an abrupt and irreversible diminishment inthe graphene's advantageous properties. According to apercolation theory, the formation of percolation islandscauses an exponential decrease in conductivity and perfor-mance [32]. In the case of CNT stretchable motors thatconnect the graphene islands to one another, however,the unique properties of the graphene surface are main-tained, even under conditions of significant deformation(Figure 2d). Therefore, the graphene flakes are able toreversibly drift with a high degree of strain while its originalproperties are maintained. A graphene sheet that is notdeformed during stretching and release offers many advan-tages for multifunctional applications. Each drifted andundeformed region of the graphene layer can be used as aconductive location for the installation of microdevices asvarious types of stretchable electronics, i.e., moleculardevices, and for stable energy storage from an electroche-mical double-layer, which was previously unachievable in astretchable system.

The stable connection between graphene flakes wasverified by monitoring the conductivity change at variousdeformed states (Figure 3). As shown in Figure 3a, theresistance of the graphene–CNT-layered structure wasnearly unchanged under a tensile strain of 80% in the case

Figure 3 Graphene–CNT-layered structure on various substrates. (a) Sheet conductivity and normalized resistance of graphene–CNT-layered structure on rubbers at different strain levels during 100 stretching–releasing cycles. (b) Differential sheet conductivityof graphene–CNT-layered structure, which shows the percolation threshold. (c) Normalized sheet conductivity of graphene–CNT-layered structure constructed on balloon during 100 expansion–deformation cycles. (d) Optical images of graphene–CNT-layeredstructures on the balloon without and with 140% stretching and shear deformation. (e) Equivalent strain, stress, and shear stress inthe x–y plane of balloon with graphene–CNT-layered structure by three-dimensional (3D)-finite element model (FEM). Black dottedboxes indicate the location of graphene–CNT layers. (f) Optical images of a light-emitting diode (LED) mounted on a surgery glovewithout and with finger bending. The images were obtained without external illumination. (g) Optical images of contact lens withLED circuit, collected with (up and left) and without (up and right) external illumination. Photographic images at different focallengths as captured by CMOS image sensor with and without the electric lens.

39Omnidirectionally stretchable, high performance supercapacitors

of uniaxial deformation, and a gradual increase in sheetconductivity was observed. The limiting strain of thestructure was 75%, as indicated by the percolation thresholdin Figure 3b. In spite of these limitations, the tendency ofconductance of the graphene–CNTstretchable electrode wasmaintained during 100 cycles of stretching–releasing with astrain of more than 150%, proving the excellent mechanicalstability and durability of the graphene–CNT electrodesystem. To examine the tolerance of the system at extre-mely mixed deformation conditions, we carried out con-ductivity measurements during the repeated contractionand expansion of a rubber balloon onto which the graphene–CNTelectrode layer was installed (Figure 3c). The expansion

of the rubber balloon by the injection of air induces asignificant simultaneous biaxial deformation and sheardeformation, which is of interest for practical applications[1–8]. Figure 3d shows the graphene–CNTelectrode region inthe contracted and expanded states during the measure-ment of conductivities. The surface extension of theinorganic muscle system on the balloon induced in thismanner reached �140% in the expanded state. In conse-quence, the electrode maintained �80% of the originalsheet conductivity, and superior electrical conductivity wasverified during more than 100 cycles. The results of 3D-FEMsimulations indicate that the extension of the graphene–CNTon the rubber balloon (the radius of the balloon was 2 cm)

I. Nam et al.40

was 136% (Figure 3e). This observation suggests that thegraphene–CNTelectrode system provides negligible mechan-ical loading on the elastic support, which is consistent withthe very low effective modulus conferred by the intrinsicproperties of the CNT clusters [16].

As a practical demonstration, we used the stretchableelectrode as an electric circuit for various stretchable electro-nics: electronic surgical gloves and hemispherical contactlenses (Figure 3f and g). The surgery glove and contact lenswere used as substrates, which shows the versatile applicationof the electrode. LED lamps in the stretchable circuits wereintegrated onto the drifting graphene flakes by conventionalsoldering. The unique mechanism of the graphene–CNT-layered system enables the integration and stretching opera-tion realized by the soldering method, which is unavailablewhen even state-of-the-art stretching circuits are used [33].Many stretchable electrodes, including electronic devices andpower supplies, have recently been reported. However, theirpractical usage is valid only for ready-made and fixedelectronic components. Most of the methods for preparingstretchable electronics demand a sophisticated procedure,because the operating components and stretching bridgesneed to be precisely separated and fixed on the surface ofthe system. However, a conventional integrating method forelectronics, represented by soldering, is inappropriate for usein such processing and soldering terminals are required for the

Figure 4 Dielectric energy storage properties of a graphene–CN(C/D) of graphene–CNT layers on PVA/H3PO4 film at a constant curspeed of 0.4% strain per minute. (b) Representative C/Ds at vastretching (80%) and twisting (3601). (c) Representative cyclic voltstrain and with stretching (80%) and twisting (3601).

connection to the electronic components. In this study, thesurface of the graphene–CNT electrode was composed of onlya graphene sheet. Therefore, the electronic devices were notin separate locations. This means that conventional electroniccomponents can be implanted on the electrode, irrespectiveof the integration method used. The LED was fixed andoperated well at stretching and bending states, which provedthat the undeformed regions supporting the device were ableto flow freely along the stretching CNTcluster. The stretchableelectrode system can also be attached to a curvilinear surface(e.g., contact lenses, without any defects), as shown inFigure 3g. The difference of sight focus with and withoutthe contact lens proves that the electric lens maintains itssight-correction ability, even after the insertion of an electriccircuit. These results show that the system provides a solutionfor the fusion of electronics and hemispherical photodetectorarrays.

The graphene flakes not only provide floating and undeform-able regions for active devices on CNTstretchable circuits, theyalso have significant electrochemical double-layer energycaused by the intrinsic high surface area (2630 m2 g�1, the-oretically) [12]. From a electrochemical double-layer mechan-ism, the specific surface area of materials is linearly propor-tional to the energy storage performance, and graphene isconsidered to be the best candidate for use in electrochemicaldouble-layer energy storage systems [10,11]. To fabricate the

T-layered structure. (a) The galvanostatic charge/discharge curverent of 0.15 A g�1 while the structure was stretched at a constantrious current densities (0.15–0.75 A g�1) without strain and withammetry curves (CVs) at various scan rates (1–10 mV s�1) without

41Omnidirectionally stretchable, high performance supercapacitors

inorganic system for a supercapacitor, we used a PVA gel as asubstrate for the graphene–CNTelectrodes [22,34]. Specifically,the PVA gel serves as (1) an elastic substrate, (2) glue, (3) aseparator, and (4) a solid solvent of electrolytic ions. Thegraphene–CNT stretchable electrodes were reversibly locatedon both sides of the PVA gel substrate. The PVA gel used herecontained H3PO4, which functions as an electrolyte. The stre-tchability of the energy storage system comes not only fromstretchable CNT clusters but also from the pseudo-elastic PVAgel substrate. The PVA gel and the CNTclusters are made up ofsimilar serpentine network chains on the atomic and micro-scale, respectively. These networks retain the constructionof the electrodes and contribute to the reversibility of def-ormation [35].

Figure 4a shows C/D curves of the supercapacitor systemwith graphene–CNT electrodes during the stretching process.The specific capacitance is 329 F g�1 based on the mass of theactive component as graphene layers, which is even higherthan the performance of conventional electrode materialswithout stretchability, as reported in previous studies [10–12].The mass of the graphene was determined by the volume ofgraphene ink spread on the substrate. A 600 μl aliquot ofgraphene ink (0.5 g/L) was used, therefore we calculated themass of the graphene to be 0.3 mg. Based on the mass of thefull cell, including passive components, the capacitance is58.4 F g�1, which is much more efficient than the capacitanceof devices fabricated using conventional techniques, includingstate-of-the-art stretchable and nonstretchable energy sto-rage systems [4–8,10–12]. The reason for these improvedresults is that the stretchable supercapacitor system isdesigned as a freestanding form and thus does not need aheavy metal current conductor, passive surface blocking activesites for stretching, or an external skeleton, which are used inconventional schemes for structure maintenance. The super-capacitor system possessed stable C/D performance for 3.3 hduring the stretching process, suggesting that the system canbe stretched up to approximately 80% strain without adecrease in performance. Figure 4b and c shows the repre-sentative C/Ds and CVs of inorganic muscles upon releasingand 80% stretching at different current and scan rates. TheCVs are rectangular in shape within a selected range ofpotential, even at high scan rates, indicating that the systemmaintained excellent capacitance behavior under highlystrained conditions. The results are also in good agreementwith the C/Ds at various current densities. Twisting providesanother substantive mode of deformation that is of interestregarding high shear deformation in the axial and widthdirections. Optical images of 3601 twisting deformations and3D-FEM simulations indicate that the system has strongmechanical stability (Figure 2b). The C/D and CV valuesalso indicate stable and high-energy storage characteristics(349 F g�1) with a twisting angle of 3601, as shown in Figure 4band c. The curves are nearly unchanged in comparison withthe original relaxed system and the performance is evenhigher, indicating the high mechanical and electrical durabilityof the system.

Conclusion

The designed graphene–CNT stretchable electrode providesimportant and extraordinary capabilities in deformable

electrodes, as demonstrated by the integration of circuitson various substrates and based on their application asenergy storage materials with significantly high perfor-mance. The suggested system, built on a 2D graphene trackwith a CNT motor, offers significant improvement overexisting methodologies for stretchable electronics in termsof stability, multimode deformation, and density of capaci-tance. We believe that our concept has noteworthy implica-tions across a number of disciplines toward the devel-opment of high-performance and stretchable electronicsfor the next generation.

Acknowledgments

This research was supported by the Global Frontier R&DProgram on Center for Multiscale Energy System funded bythe National Research Foundation under the Ministry ofScience, ICT & Future, Korea (NRF-2011-0031571) and theSupercomputing Center/Korea Institute of Science andTechnology Information with supercomputing resourcesincluding technical support (KSC-2014-C1-013).

Appendix A. Supporting information

Supplementary data associated with this article can befound in the online version at http://dx.doi.org/10.1016/j.nanoen.2015.04.001.

References

[1] J.A. Rogers, T. Someya, Y. Huang, Science 327 (2009)1603–1607.

[2] S. Xu, Y. Zhang, L. Jia, K.E. Mathewson, K.-I. Jang, J. Kim,H. Fu, X. Huang, P. Chava, R. Wang, S. Bhole, L. Wang, Y.J. Na,Y. Guan, M. Flavin, Z. Han, Y. Huang, J.A. Rogers, Science 344(2014) 70–74.

[3] K.S. Kim, Y. Zhao, H. Jang, S.Y. Lee, J.M. Kim, K.S. Kim,J.-H. Ahn, P. Kim, J.-Y. Choi, B.H. Hong, Nature 457 (2009)706–710.

[4] D. Kim, G. Shin, Y.J. Kang, W. Kim, J.S. Ha, ACS Nano 7 (2013)7975–7982.

[5] S. Xu, Y. Zhang, J. Cho, J. Lee, X. Huang, L. Jia, J.A. Fan,Y. Su, J. Su, H. Zhang, H. Cheng, B. Lu, C. Yu, C. Chuang,T. Kim, T. Song, K. Shigeta, S. Kang, C. Dagderviren, I. Petrov,P.V. Braun, Y. Huang, U. Paik, J.A. Rogers, Nat. Commun. 4(2013) 1543.

[6] X. Li, T. Gu, B. Wei, Nano Lett. 12 (2012) 6336–6371.[7] Z. Niu, H. Dong, B. Zhu, J. Li, H.H. Hng, W. Zhou, X. Chen,

S. Xie, Adv. Mater. 25 (2013) 1058–1064.[8] L. Hu, M. Pasta, F.L. Mantia, L. Cui, S. Jeong, H.D. Deshazer,

J.W. Choi, S.M. Han, Y. Cui., Nano Lett. 10 (2010) 708–714.[9] M.J. Allen, V.C. Tung, R.B. Kaner, Chem. Rev. 110 (2010)

132–145.[10] J.H. Lee, N. Park, B.G. Kim, D.S. Jung, K. Im, J. Hur,

J.W. Choi, ACS Nano 7 (2013) 9366–9374.[11] J.J. Yoo, K. Balakrishnan, J. Huang, V. Meunier, B.G. Sumpter,

A. Srivastava, M. Conway, A.L.M. Reddy, J. Yu, R. Vajtai,P.M. Ajayan, Nano Lett. 11 (2011) 1423–1427.

[12] Y. Zhu, S. Murali, M.D. Stoller, K.J. Ganesh, W. Cai,P.J. Rerreira, A. Pirkle, R.M. Wallace, K.A. Cychosz,M. Thommes, D. Su, E.A. Stach, R.S. Ruoff, Science 332(2011) 1537–1541.

I. Nam et al.42

[13] S. Won, Y. Hwangbo, S.-K. Lee, K.-S. Kim, K.-S. Kim, S.-M. Lee,H.-J. Lee, J.-H. Ahn, J.-H. Kim, S.-B. Lee, Nanoscale 6 (2014)6057–6064.

[14] J.M. Berg, J.L. Tymoczko, L. Stryer, Biochemistry7th ed.,Palgrave Macmillan, New York, USA, 2012.

[15] M. Kneussel, W. Wagner, Nat. Rev. 14 (2013) 233–247.[16] N.J. Ginga, W. Chen, S.K. Sitaraman, Carbon 66 (2014) 57–66.[17] Y. Won, Y. Gao, M.A. Panzer, R. Xiang, S. Maruyama,

T.M. Kenny, W. Cai, K.E. Goodson, Proc. Natl. Acad. Sci. USA110 (2013) 20426–20430.

[18] F. Liu, P. Ming, J. Li, Phys. Rev. B 76 (2007) 064120.[19] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996)

3865–3868.[20] G. Kresse, J. Furthmüller, Comput. Mater. Sci. 6 (1996) 15–50.[21] S.J. Grimme, Comput. Chem. 27 (2006) 1787–1799.[22] I. Nam, G.-P. Kim, S. Park, J.W. Han, J. Yi, Energy Environ. Sci.

7 (2014) 1095–1102.[23] Y. Zhu, S. Murali, M.D. Stoller, A. Velamakanni, R.D. Piner,

R.S. Ruoff, Carbon 48 (2010) 2106–2122.[24] I. Nam, S. Park, G.-P. Kim, J. Park, J. Yi, Chem. Sci. 4 (2013)

1663–1667.[25] M. Seydou, S. Marsaudon, J. Buchoux, J.P. Aimé, A.M. Bonnot,

Phys. Rev. B 80 (2009) 245421.[26] O.V. Kibis, D.G.W. Parfitt, M.E. Portnoi, Phys. Rev. B 71 (2005)

035411.[27] A. Fasolino, J.H. Los, M.I. Katsnelson, Nat. Mater. 6 (2007)

858.[28] B.G. Demczyk, Y.M. Wang, J. Cumings, M. Hetman, W. Han,

A. Zettl, R.O. Ritchie, Mater. Sci. Eng. A334 (2002) 173.[29] D.C. Hammerand, G.D. Seidel, D.C. Lagoudas, Mech. Adv.

Mater. Struct. 14 (2007) 277–294.[30] C. Lee, X. Wei, J.W. Kysar, J. Hone, Science 321 (2008) 385.[31] J.-W. Jiang, J.-S. Wang, B. Li, Phys. Rev. B 80 (2009) 113405.[32] P.J. King, T.M. Higgins, S. De, N. Nicoloso, J.N. Coleman, ACS

nano 6 (2012) 1732–1741.[33] G.S. Jeong, D.-H. Baek, H.C. Jung, J.H. Song, J.H. Moon,

S.W. Hong, I.Y. Kim, S.-H. Lee, Nat. Commun. 3 (2012) 977.[34] M. Kaempgen, C.K. Chan, J. Ma, Y. Cui, G. Gruner, Nano Lett.

9 (2009) 1872–1876.[35] W. Chen, X. Tao, P. Xue, X. Cheng, Appl. Surf. Sci. 252 (2005)

1404–1409.

Inho Nam received his B.S. degree fromYonsei University in 2010. He is currentlyworking with Professor Jongheop Yi as aPh.D candidate in School of Chemical andBiological Engineering at Seoul NationalUniversity under Academic Future Genera-tions Fellowship. His research focuses onnanostructured materials and stretchabledevices for advanced electronics.

Seongjun Bae graduated the School of Che-mical and Biological Engineering at SeoulNational University, Republic of Korea in2013. Now he is a graduate student of Schoolof Chemical and Biological Engineering fromthe Seoul National University since 2013. Heis studying for a Ph. D. course in Prof.Jongheop Yi's group. He is currently research-ing on the field of energy storage system.

Soomin Park is a Ph.D. candidate in the Schoolof Chemical and Biological Engineering of SeoulNational University. He received a B.S. degree(2010) from the Department of Biotechnologyof Yonsei University. His chief interests includethe rational design and experimental assess-ment of advanced electrode materials forenergy storage systems such as Li/Na recharge-able batteries and supercapacitors.

Young Geun Yoo received his B.S. degree inchemistry at Yonsei University, Republic ofKorea in 2013. Now he is a Ph.D. candidateof School of Chemical and Biological Engi-neering in Seoul National University since2014. He is currently researching on energystorage system and electrochemical engi-neering under the supervision of Prof.Jongheop Yi.

Jong Min Lee is an associated professor ofSchool of Chemical and Biological Engineer-ing at Seoul National University. Afterreceiving his Ph.D. degree in ChemicalEngineering in 2004 from Georgia Instituteof Technology, he worked as a post-doctoralfellow at Georgia Institute of Technologyand the University of Virginia. He thenserved as an assistant professor of Depart-ment of Chemical and Materials Engineering

at the University of Alberta from 2006 to 2010. In 2010, he joinedthe faculty of the SNU and his research interests focuses on theadvanced chemical process modeling and control.

Jeong Woo Han received his B.S. degree inChemical and Biological Engineering fromthe Seoul National University in 2005 and hisPh.D. degree in Chemical and BiomolecularEngineering from Georgia Institute of Tech-nology in 2010. Following a post-doctoralfellow at Massachusetts Institute of Tech-nology, he became an assistant professor inDepartment of Chemical Engineering at theUniversity of Seoul in 2012. His research

interests focuses on the computational design of energy materialsand catalysis.

Jongheop Yi received his B.S. degree inChemical Engineering from the SeoulNational University in 1980 and his Ph.D.degree in Chemical Engineering from Syra-cuse University in 1991. He went on to workas a project leader at Rhone-Poulenc inHouston. In 1993, he became a professorin the School of Chemical and BiologicalEngineering at Seoul National University. Heleads a group of researchers working on

nano-materials and processes for energy and environmentapplications.