Hybrid Integration of Carbon Nanotubes and Transition MetalDichalcogenides on Cellulose Paper for Highly Sensitive andExtremely Deformable Chemical SensorsWoo Sung Lee and Jungwook Choi*

School of Mechanical Engineering, Yeungnam University, 280 Daehak-ro, Gyeongsan, Gyeongbuk 38541, Republic of Korea

*S Supporting Information

ABSTRACT: Sensitive and deformable chemical sensors manu-factured by a low-cost process are promising as they aredisposable, can be applied on curved, complex structures, andprovide environmental information to users. Although manynanomaterial-based flexible sensors have been suggested to meetthese demands, their limited chemical sensitivity and mechanicalflexibility pose challenges. Here, a highly deformable chemicalsensor is reported with improved sensitivity that integratesmultiwalled carbon nanotubes (CNTs) and nanolayered tran-sition metal dichalcogenides (TMDCs) on cellulose paper. Liquiddispersions of CNTs and TMDCs are absorbed and dried onporous cellulose for sensor fabrication, which is simple, scalable,rapid, and inexpensive. The cellulose substrate enables reversiblethree-dimensional folding and unfolding, bending down to 0.25 mm, and twisting up to 1800° (∼628.4 rad m−1) withoutdegradation, and the CNTs maintain a percolation network and simultaneously provide gas reactivity. Functionalization ofCNTs with TMDCs (WS2 or MoS2) greatly improves the sensing response upon exposure to NO2 molecules by more than150%, and the sensor can also selectively detect NO2 over diverse reducing vapors. The measured NO2 sensitivity is 4.57%ppm−1, which is much higher than that of previous paper-based sensors. Our sensor can stably and sensitively detect the gaseven under severe deformation such as heavy folding and crumpling. Hybrid integration of CNTs and TMDCs on cellulosepaper may also be used to detect other harmful gases and can be applicable in low-cost portable devices that require reliabledeformability.

KEYWORDS: carbon nanotube, transition metal dichalcogenide, cellulose substrate, deformable device, sensor-on-paper

■ INTRODUCTION

Recent advances in materials and manufacturing processeshave enabled the realization of soft electronics, optoelectronics,and sensors with high performance comparable to that ofconventional rigid devices.1 Various organic and inorganicmaterials with micro/nanoscale dimensions have beenintegrated onto elastomeric substrates for electronic skin, softrobotics, flexible displays, energy storage, and wearablesensors.2,3 These devices can be mounted on nonplanarsurfaces and exhibit compelling performance; however, theyusually require sophisticated fabrication processes and havelimited flexibility.4 Considering the increasing demand for low-cost and highly deformable devices,5,6 achieving reliablefoldability and twistability would make it possible to improvethe functionality of devices and to provide new uses for themfor integration into more complex components.Cellulose paper is an alternative that can overcome the

limitations of conventional flexible substrates. Owing to its lowcost, lightweight, disposability, biodegradability, and deform-ability, various applications ranging from flexible electronics toelectrochemical devices have been successfully demonstra-

ted.4,5,7−10 In addition, because of its unique features, cellulosepaper can serve as a scaffold for printing, filtration, anddeposition of functional materials for flexible sensors.11−17 Inparticular, integration of nanomaterials on cellulose paper canbe an ideal approach to flexible chemical sensors because oftheir porosity, large surface area, and bendability, asdemonstrated by carbon nanotubes (CNTs),18−23 gra-phene,24,25 and nanoparticles.26−28 Among many nanoscalematerials and structures, carbon nanomaterials have attractedsignificant interest as sensing materials because of theirexcellent material properties.29 Carbon nanomaterials inte-grated on paper offer flexibility and responsiveness to diversechemical species; however, their sensitivity to harmful gasessuch as NO2 is less than ∼1.5% ppm−1,22,24,25 which is lowerthan that of other nanostructured flexible sensors. Thus, it isnecessary to develop gas sensors having high chemical

Received: February 21, 2019Accepted: May 7, 2019Published: May 7, 2019

Research Article

www.acsami.orgCite This: ACS Appl. Mater. Interfaces 2019, 11, 19363−19371

© 2019 American Chemical Society 19363 DOI: 10.1021/acsami.9b03296ACS Appl. Mater. Interfaces 2019, 11, 19363−19371

Dow

nloa

ded

by X

IAN

GT

AN

UN

IV a

t 07:

59:2

7:06

4 on

Jun

e 03

, 201

9fr

om h

ttps:

//pub

s.ac

s.or

g/do

i/10.

1021

/acs

ami.9

b032

96.

sensitivity while exploiting the advantages of the papersubstrate.Here, we present the first use of a paper-based, extremely

deformable gas sensor using integrated one-dimensional (1D)CNTs and two-dimensional (2D) transition metal dichalcoge-nides (TMDCs) to improve the chemical sensitivity. As surfacefunctionalization using heterogeneous, multidimensional ma-terials improves the sensing response,30−32 1D CNT networkson cellulose paper were decorated with 2D TMDCs.Semiconducting TMDCs have attracted great interest owingto not only their layer-number-dependent properties,33 butalso their high physical and chemical reactivity.34,35 Thus, thehigh chemical sensitivity of CNTs and TMDCs and chargetransfer between them upon adsorption of chemical species areexpected to afford significantly improved sensitivity, and thecellulose substrate provides high deformability.In this work, we demonstrate a highly sensitive, foldable, and

twistable chemical sensor based on a hybrid of 1D CNTs and2D TMDCs (WS2 or MoS2) on cellulose paper. Fabricationrequires only absorption and evaporation of CNT and TMDCdispersions on cellulose, so it is simple, scalable, rapid (within30 min), and inexpensive (∼$0.23 per sensor for chemicals andmaterials). Cellulose paper provides the porosity, three-dimensional (3D) foldability, and extreme twistability neededfor robust flexible sensors, and the CNTs offer a percolationpathway and gas sensitivity. When CNTs on cellulose arefunctionalized with WS2, the response to 10 ppm NO2 isimproved by ∼150%. The versatility of our approach is furtherdemonstrated by functionalizing CNTs with MoS2 or bydetecting other chemical species, in which case the sensingbehavior is opposite to that of NO2. In addition, the sensor canbe reversibly bent (down to a 0.25 mm bending radius) andtwisted (up to 1800°) with high cyclic durability and can stablyand sensitively detect the gas even under heavy crumpling and

creasing. Our sensor not only outperforms previously reportedlow-sensitive paper-based chemical sensors but also provideshigh stability against severe mechanical deformation.

■ RESULTS AND DISCUSSIONFigure 1a schematically illustrates the fabrication of a cellulose-paper-based CNT−TMDC hybrid sensor. CNT and TMDCpowders were sonicated, centrifuged, and finally stablydispersed in N,N-dimethylformamide (DMF) at concentra-tions of ∼2 and ∼0.2 mg mL−1, respectively (see theExperimental Section for details on material preparation).We used multiwalled CNTs (outer diameter, ∼15 nm) for allof the experiments and WS2 nanolayers (number of layers, twoto three) as a representative TMDC; the number of WS2nanolayers was confirmed by transmission electron microscopy(TEM) (Figure S1, Supporting Information). The UV-visibleabsorption spectrum of the WS2 dispersion clearly indicatedlayered WS2 with excitonic signatures at 463, 538, and 642 nm,which correspond to those in a previous report (Figure S2a,b,Supporting information).36 The CNT dispersion showed anabsorption peak around 272 nm (Figure S2c,d, SupportingInformation). Pristine cellulose paper was repeatedly dip-coated in the CNT dispersion and dried until a percolationnetwork formed and the measurable electrical resistance wasobtained; the coated paper was then dip-coated in the WS2dispersion and dried. As a result of repetitive coating processes,both CNTs and WS2 were bound on the surface of cellulosepaper possibly by van der Waals interactions. An elasticsubstrate such as poly(dimethylsiloxane) (PDMS) can option-ally be used to support the nanomaterial-coated cellulosepaper. The entire fabrication process takes only ∼30 min,including dip-coating, drying, and wiring electrical connec-tions. Owing to the process scalability, nanomaterial-coatedlarge-area paper more than 100 cm2 in size can be prepared

Figure 1. Multidimensional, heterogeneous integration of 1D CNTs and 2D TMDCs on cellulose paper. (a) Schematic of the preparation processof CNT−TMDC-integrated cellulose paper for a sensitive and deformable chemical sensor. The cellulose paper is sequentially dip-coated in CNTand TMDC dispersions and dried until a percolation network forms. The entire process is simple and rapid, so it is suitable for producing low-costdisposable sensors. The nanomaterial-coated paper is self-standing, and an elastic substrate can optionally be used as a support or protection layer.(b) Photograph of nanomaterial-coated large-area (>100 cm2) cellulose paper, showing process scalability. The paper can be easily cut into variousshapes and sizes. (c) 3D origami boat folded from the nanomaterial-coated paper to connect an LED circuit. (d) Photographs of the bent andtwisted cellulose paper mounted on a PDMS substrate, demonstrating its deformability.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.9b03296ACS Appl. Mater. Interfaces 2019, 11, 19363−19371

19364

(Figure 1b), and it can be scaled up simply by increasing thequantity of dispersed nanomaterials and the size of thecellulose paper. Because of the paper substrate, the sensor canbe creased, folded, and unfolded to form 3D self-standingstructures; if necessary, it can be trimmed and perforatedeasily. Figure 1c shows a 3D origami boat folded from thenanomaterial-coated paper, which can be used to connect alight-emitting diode (LED) circuit. The deformability of thecellulose paper under bending and twisting is shown in Figure1d.The scanning electron microscopy (SEM) images in Figure

2a,b show the cellulose microfibers, on which the CNTs formpercolation pathways, and WS2 nanolayers locally deposited onthem. The cellulose microfibers maintain their network aftermultiple dip-coating processes until the CNTs completelycover their surfaces (Figure S3, Supporting Information).Raman spectra of the cellulose paper before and after CNTand WS2 coating were collected under 532 nm excitation(Figure 2c). Distinct Raman signatures of multiwalled CNTswere fitted into four Lorentzian curves, which can be assignedto D, D″, G, and D′ peaks located at ∼1350, ∼1469, ∼1577,and 1612 cm−1 (red line in Figure 2c). The intensity ratio of Dand G peaks of the CNTs before and after WS2 coating is∼0.96 and ∼1.05, respectively. The comparable intensity of theD and G peaks indicates the presence of defects and disorder

in the sp2-hybridized carbon structure, which may haveoriginated during sonication and dispersion of the CNTs.Note that these defective CNTs could be useful for improvingthe gas sensitivity, as imperfections provide chemisorption sitesfor NO2 molecules.37 In the Raman spectrum measured afterWS2 coating (blue line in Figure 2c), the distinctive Ramansignatures of WS2 at 350 cm−1 (in-plane mode, E2g

1) and 418cm−1 (out-of-plane mode, A1g) emerge over the cellulose peaks(Figure 2d). The difference in peak position and the intensityratio of E2g

1 and A1g are ∼68 cm−1 and ∼1.29, respectively,indicating that there are two to three layers of WS2.

38 TheRaman spectra clearly demonstrate the successful integrationof CNTs and WS2 on the cellulose paper.We also characterized the elemental composition of the

CNT−WS2-coated cellulose paper using X-ray photoelectronspectroscopy (XPS), as shown in Figure 2e−g. The core-levelC 1s spectrum of the CNTs (Figure 2e) was deconvoluted intomultiple peaks. The major peak, which is centered at 284.4 eV,originates from sp2-hybridized carbon. Other peaks at 285.2,285.9, 287.4, and 291.0 eV can be assigned to sp3 bonding, C−O−, CO, and −COO−, respectively, in agreement withprevious studies.39 The sp3 and oxidized carbon structures canbe considered as defect sites that may promote molecularadsorption; this result is also consistent with the presence ofthe D peak in the Raman spectrum. The W 4f core-level XPS

Figure 2. Characterization of the nanomaterial-integrated cellulose paper. (a, b) SEM images of the 1D CNT-2D WS2 on cellulose microfibers. TheCNTs completely cover the surfaces of the cellulose fibers, forming percolation pathways, and the WS2 is deposited locally on them. An enlargedSEM image of a red box in (a) is shown in (b). (c, d) Raman spectra of pristine cellulose (green line), CNT-coated cellulose (red line), and CNT−WS2-coated cellulose (blue line). Raman spectra collected from the CNTs are deconvoluted into four curves. The CNTs exhibit D and G peaks(∼1350 and ∼1577 cm−1, respectively), and WS2 displays in-plane (E2g

1) and out-of-plane (A1g) modes at 350 and 418 cm−1, respectively.Magnified Raman spectra in (d) are from the region enclosed by a dotted line in (c). (e−g) X-ray photoelectron spectra (XPS) of CNT−WS2-coated cellulose. The C 1s (e), W 4f (f), and S 2p (g) core-level XPS spectra indicate successful integration of CNTs and WS2 without significantdegradation during sonication, dispersion, coating, and drying.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.9b03296ACS Appl. Mater. Interfaces 2019, 11, 19363−19371

19365

spectrum has a 4f7/2 and 4f5/2 doublet at binding energies of∼32.68 and 34.98 eV, respectively (Figure 2f). In addition, thebinding energies of S 2p3/2 and 2p1/2 are approximately 162.38and 163.58 eV (Figure 2g). The XPS spectra of WS2 depositedon the CNTs show a slight blue shift compared to that in aprevious report on pristine WS2.

31 As the shift in bindingenergy is correlated with a change in the Fermi level, this blueshift can originate from n-type doping (an increase in theelectron concentration) of WS2.

40 The reason could be holetransfer from WS2 to the CNTs driven by the Fermi leveldifference between them, which is further verified bymeasuring the resistance change of the CNTs as a functionof the number of WS2 coatings.Increasing the number of CNT coatings dramatically

decreases the resistance of the cellulose paper (Figure 3a).The initial resistance was not measurable (>1 GΩ), and itbecame saturated down to ∼10.7 kΩ after several cycles ofCNT coating and drying. As DMF, which we used as a solventto disperse the nanomaterials, could interact with the CNTs,the CNT-coated cellulose paper was immersed in pure DMFand dried as a control experiment. Possibly owing to theinteraction between the nanomaterial and solvent,41,42 theresistance of the CNTs increased as the number of interactionswith DMF increased (Figure S4a, Supporting Information).However, when the CNT-coated cellulose was immersed anddried in the WS2−DMF dispersion, the increasing trend of theresistance was suppressed, indicating that the locally depositedWS2 decreased the resistance of the CNT network (FigureS4b,c, Supporting Information). Thus, interaction of the CNTswith DMF can be excluded. The resistance change is plottedversus the number of WS2 coatings in Figure 3b. CNTs andWS2 reportedly behave as a p-type semiconductor whenexposed to ambient air,43 and as shown by the XPS studies, theFermi level difference between CNT and WS2 could transferholes from WS2 to CNTs, increasing the number of carriers inthe CNTs and ultimately decreasing the resistance (Figure S5,Supporting Information). The insets in Figure 3a,b show thecorresponding linear I−V curves with respect to the number ofnanomaterial coatings. Note that the fabrication process ishighly reproducible, as there is minimal sample-to-samplevariation (Figure S6, Supporting Information).The transient gas sensing response of the CNT−cellulose

and CNT−WS2−cellulose was investigated under exposure toNO2 with air as a carrier gas. The response is defined as |ΔR/

R0|, where ΔR is the difference between the resistance beforeand after exposure to NO2, and R0 is the initial resistance.Although the presence of signal drift and the response variationbetween subsequent exposures are observed, the sensingresponse between CNTs and CNT−WS2 is clearly distinguish-able. As shown in Figure 4a, the response is improved by morethan 150% after the CNTs are functionalized with WS2compared to that of the CNTs alone under identical, repeated10 ppm NO2 exposure in air. This improvement may resultfrom the high reactivity of WS2 and the increased number ofcarriers in the CNTs owing to charge transfer from the NO2-adsorbed WS2 (Figure S7, Supporting Information). Toelucidate the effect of WS2 decoration, we compared theresponses of three sensors that were prepared with five CNTcoatings, 10 CNT coatings, and five CNT coatings with anadditional five WS2 coatings. Increasing the number of CNTcoatings did not further enhance the gas sensing response, butthe addition of WS2 to the CNTs noticeably improved theresponse, experimentally demonstrating the utility of WS2(Figure S8a,b, Supporting Information).Figure 4b,c shows the sensing response of the CNT−WS2-

integrated cellulose paper to NO2 concentrations ranging from0.1 to 10 ppm. The response increases with increasing NO2concentration owing to an increase in the number of adsorbedmolecules on the CNTs and WS2. However, signal drift isobserved, which can be attributed to the high adsorptionenergy of NO2 on CNTs (∼0.7 eV) and WS2 (∼0.4 eV).44,45

This incomplete desorption of NO2 molecules is amelioratedby UV light irradiation during the recovery cycles (Figure S9,Supporting Information). Otherwise, additional functionaliza-tion of WS2 with silver nanowires would improve the recoveryafter NO2 adsorption.

31 As shown in Figure 4d, the sensitivity(defined as the response per ppm NO2) reaches 4.57% ppm−1

at low concentrations (<2 ppm), which is ∼3−22 times higherthan those of previously reported NO2 sensors22,24,25 (seeTable S1, Supporting Information). Although some nanoma-terial-based flexible sensors have exhibited higher sensitivity, tothe best of our knowledge, our sensor shows the highestsensitivity among the reported NO2 sensors based on carbonnanomaterials on paper. At high concentrations (>2 ppm), thesensitivity decreases to 0.84% ppm−1 and gradually becomessaturated as the NO2 concentration increases. Instead ofinterpretation by bilinear sensitivity, entire responses as afunction of NO2 concentration are also well fitted to the

Figure 3. Electrical characterization of cellulose paper (100 mm2) with respect to number of (a) CNT and (b) WS2 coatings. Increasing thenumber of CNT coatings on the cellulose paper dramatically decreases the resistance, and the CNT network serves as both an electrical conductorand a chemiresistor. With three CNT coatings, the resistance becomes measurable (1.625 MΩ), and it finally saturates down to 10.7 kΩ. Increasingthe number of WS2 coatings on the CNT-coated cellulose paper also decreases the resistance. Note that the effect of the dispersing solvent (DMF)on the resistance of the CNTs was excluded. Locally deposited WS2 enabled charge transfer to the CNTs, resulting in a decrease in the resistance to2.5 kΩ. Insets in (a) and (b) show the corresponding linear I−V curves.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.9b03296ACS Appl. Mater. Interfaces 2019, 11, 19363−19371

19366

Langmuir adsorption model.46 This implies that the responseof our sensor is dominated by the amount of NO2 adsorptionat the surface of CNTs and WS2. The high sensitivity at lowconcentrations would be useful for early detection of NO2. Inaddition, because of its low fabrication cost and low laborintensity, our sensor can be considered as a disposable devicewith advanced sensitivity.The CNT−WS2-integrated cellulose paper can selectively

detect NO2 over diverse chemical species such as acetone,ethanol, p-xylene, toluene, ethylbenzene, NH3, and CO (Figure5a−h). All experiments were conducted in an air environment,

and the concentration of all target gases was 10 ppm. Thetested chemical vapors, except oxidizing gas such as NO2, areknown to be a reducing gas that donates electrons to sensingmaterials.47,48 As such, the resistance of the sensor increasedupon exposure to them unlike NO2 sensing. Among the testedgases, the sensor exhibits the highest response to NO2 underthe identical experimental conditions, as shown in Figure 5i.This selectivity would be a result from the differences inadsorption energy between the sensing material and adsorbedmolecules and the corresponding charge transfer betweenthem.49,50 Thus, this opposite and relatively small resistancechange toward other gases would be useful for selectiveidentification of low NO2 concentration.

Figure 4. NO2 sensing response of CNT−WS2−cellulose paper. (a)Transient sensing response of CNT- and CNT−WS2-coated paperunder repeated exposure to 10 ppm NO2 in an air environment. Theresistance decreases upon exposure to NO2 in both cases, andfunctionalization of the CNTs with WS2 greatly improves theresponse (|ΔR/R0|), by more than 150%. (b, c) NO2-concentration-dependent response of the CNT−WS2 hybrid from 0.1 to 0.6 ppm(b) and from 1 to 10 ppm (c). The response increases in proportionto the NO2 concentration. (d) Sensitivity (response per ppm) of thesensor, which is well fitted to the Langmuir adsorption model (redline). Alternatively, the sensitivity can also be interpreted as two linearregions at 0.1−2 ppm (R2 = 0.985) and 2−10 ppm (R2 = 0.953) (bluelines). The high sensitivity (4.57% ppm−1) at low NO2 concentrations(<2 ppm) would be advantageous for early detection of NO2molecules.

Figure 5. Selectivity of the CNT−WS2 hybrid sensor. (a−h)Responses of the sensor under exposure to diverse chemical speciessuch as acetone (a), ethanol (b), p-xylene (c), toluene (d),ethylbenzene (e), NH3 (f), CO (g), and NO2 (h). All experimentalconditions are identical including gas concentration (10 ppm), carriergas (air), exposure and recovery time, and applied bias voltage (1 V),except the species of gas molecules. (i) Comparison of sensingresponses toward the chemical vapors tested. The sensor shows thehighest response to NO2 compared to the other gases, demonstratingits high selectivity. It is noted that the resistance of the sensorincreases for all other gases except NO2.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.9b03296ACS Appl. Mater. Interfaces 2019, 11, 19363−19371

19367

In addition, our approach is not limited to the functionaliza-tion of CNTs with WS2 but can be extended to decorationwith other TMDCs such as MoS2. We also exfoliated MoS2 inDMF and measured its absorption spectrum, confirming thestable dispersion of MoS2 nanolayers with distinctive excitonsignatures (Figure S10a,b, Supporting Information). After theMoS2 was deposited on the CNT−cellulose, Raman spectros-copy verified the successful integration, and the NO2 responseof CNT−MoS2 (∼12% response at 10 ppm NO2), which wassimilar to that of CNT−WS2 but higher than that of CNTsalone, was investigated (Figure S10c,d, Supporting Informa-tion).Of considerable utility of the cellulose substrate is extreme,

reversible deformability and durability. To verify thedeformability of the sensor, the resistance was recordedunder twisting as well as concave and convex bending. Asshown in Figure 6a, the relative resistance change of the sensorwas negligible and reversibly returned to its initial state forboth concave and convex bending down to a bending radius of0.25 mm. This high bendability illustrates the outstandingfeature of cellulose paper, as other soft substrates (which areusually difficult to handle owing to high compliance) generallyneed to be sufficiently thin to achieve such a low bendingradius.51 We also investigated the reversible twistability of oursensor up to 1800° (∼628.4 rad m−1), and no degradation wasobserved, as shown in Figure 6b. Unlike the negligibleresistance changes under bending, the resistance is slightlysmaller under twisting than in the untwisted flat state owing toan increase in the contact area of the cellulose paper. Inaddition, the insignificant variation of the resistance over 1000cycles of bending and twisting at 0.85 mm radius and 1080°,respectively, also demonstrates the high mechanical deform-ability and electrical reliability of the sensor (Figure 6c).

Finally, owing to its deformability, the sensor can be stablyoperated even under heavy folding and crumpling withoutcompromising the sensitivity, as shown in Figure 6d. Theresponse to 10 ppm NO2 exposure is 14.15, 17.23, and 16.1%under bending, crumpling, and folding, respectively, which iscomparable to or even better than that in the undeformed, flatstate (responsiveness ∼ 14.3%).

■ CONCLUSIONS

We demonstrated cellulose-paper-based chemical sensors using1D multiwalled CNTs and 2D nanolayered TMDCs as sensingmaterials. Heterogeneous integration of TMDCs on CNTssignificantly improved the chemical reactivity compared to thatof the homogeneous CNTs, thus avoiding the low sensitivity ofprevious paper-based sensors. At low NO2 concentrationbelow 2 ppm, the sensor exhibits high sensitivity of 4.57%ppm−1. Furthermore, the sensing response to NO2 was thehighest among other tested vapors, demonstrating the highselectivity of our sensor. At the same time, the cellulosesubstrate allowed extreme but reliable mechanical deformation,as verified by folding in self-standing 3D origami, reversiblebending (down to 0.25 mm) and twisting (up to 1800°), andchemical sensing under crumpling. Given the scalability,simplicity, and low cost of the fabrication process, the materialcould be used in disposable environmental sensors that can beattached to curved and complicated surfaces. Our approachcould be extended to various paper- or fabric-based devices bycombining diverse materials that can be dispersed in liquids.Examples may range from a stretchable strain sensor withmesh-structured paper fabricated by trimming using scissors toa flexible photodetector that uses the excitonic transitions ofsemiconducting TMDCs under visible light. In addition, thedeformation-insensitive resistance change of the nanomaterial-

Figure 6. Deformability of the CNT−WS2 hybrid sensor on cellulose paper. (a) Concave and convex bending from the initial flat state to a bendingradius of 0.25 mm. The resistance change is negligible and reversible. (b) Twisting up to 1800° (∼628.4 rad m−1). The resistance change is alsoreversible, and no degradation is observed. Both bending (a) and twisting (b) tests demonstrate the extreme deformability and durability of thesensor. (c) Cyclic durability against repeated bending (0.85 mm bending radius) and twisting (1080°) up to 1000 cycles. The resistance change ofthe sensor is insignificant (<0.25%) in both cases. (d) Detection of 10 ppm NO2 under heavy deformation. The high deformability of both celluloseand nanomaterials and the improved chemical sensitivity result in stable and sensitive detection, with a response of 14.3, 14.15, 17.23, and 16.1% inthe flat state and under bending, crumpling, and folding, respectively. Insets show photographs of the deformed sensors used for the experiments.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.9b03296ACS Appl. Mater. Interfaces 2019, 11, 19363−19371

19368

coated cellulose demonstrated in this work would beadvantageous for decoupling signals from other stimuli thanbending or twisting.

■ EXPERIMENTAL SECTIONPreparation of CNT and TMDC Dispersions.Multiwalled CNT

powder was purchased from Graphene Supermarket and used asreceived. The CNTs were dispersed in DMF (99.5%, DaejungChemicals) using a tip sonicator (HD4100, Bandelin) with a power of100 W, an amplitude of 30%, and a frequency of 20 kHz for 30 min atice-cooled temperature. WS2 and MoS2 powders were purchased fromGraphene Supermarket and used as received. They were exfoliatedand also dispersed in DMF by a process similar to that of the CNTs,by sonication at 100 W, an amplitude of 30%, and 20 kHz for 120 minat ice-cooled temperature. To avoid overheating during sonication,the duty cycle was set to 20%. All of the nanomaterial dispersionswere centrifuged at 6000 rpm for 30 min using a Hettich Rotina 380centrifuge; then, the top 10% of the supernatant was collected forfurther use. The concentrations of the CNT and TMDC dispersionswere approximately 2 and 0.2 mg mL−1, respectively. The finaldispersions were held under atmospheric conditions at roomtemperature.Fabrication of CNT- and TMDC-Coated Cellulose Paper.

Cellulose paper (Kimwipes, Kimberly-Clark) was dip-coated in theCNT dispersion for 5 s and then dried at 120 °C for 120 s. After theCNTs formed a percolation network and the desired resistance wasreached, the CNT-coated cellulose was dip-coated in the TMDCdispersion by a procedure similar to that used for CNT coating.PDMS (Sylgard 184, Dow Corning), which can optionally be used tosupport the cellulose paper, was prepared by mixing the base andcuring agent at a weight ratio of 10:1 and subsequent curing at 150 °Cfor 30 min. Silver paste (Elcoat P-100, CANS) was used for electricalwiring on the nanomaterial-coated cellulose paper.Characterization of the Materials and Sensor. A PerkinElmer

Lambda 950 UV/vis/near-IR spectrophotometer was used to collectthe absorption spectra of the CNT and TMDC dispersions. Ramanspectra were recorded using a Raman microscope (XploRA Plus,Horiba) using a 532 nm laser and a 100× objective lens at roomtemperature. The spot diameter of the laser was ∼1 μm, and we useda low laser power to avoid overheating or damaging the samples. Thenanomaterials were characterized by field-emission TEM (H-7600,Hitachi), and the surface morphology of the sensors was inspected byfield-emission SEM (S-4800, Hitachi). XPS measurements wereperformed using a Kα XPS system (Thermo Scientific) with amonochromatic Al Kα X-ray source. The XPS spectra weredeconvoluted into Gaussian curves for C 1s fitting and Lorentziancurves for W 4f and S 2p fitting. To test the electrical connectivity ofthe 3D folded origami boat, a direct current power supply (2220G,Keithley) was used. For the bending and twisting experiments on thesensor, custom-built translational and rotational stages were used, andthe resistance change was recorded using a digital multimeter (2000,Keithley).Gas Sensing Experiment. A tightly sealed chamber composed of

a gas inlet, an outlet, a temperature controller, and an electricalfeedthrough was used for the gas sensing test. We used pure air, air-balanced 10 ppm NO2, acetone, ethanol, p-xylene, toluene, ethyl-benzene, NH3, and CO, and a constant total flow rate of 500 sccmwas maintained during all of the experiments by mass flow ratecontrollers. The environmental temperature during the sensing testwas modulated from 25 to 150 °C (see Figure S11, SupportingInformation, for the temperature coefficient of the resistance of thesensor). To clearly observe the difference in sensing response betweenthe CNTs and CNT−TMDC samples, sensing was performed at 150°C, unless specified otherwise. Note that the sensing behavior wasalso similar at room temperature (Figure S12a, SupportingInformation). In addition, UV irradiation facilitated the NO2desorption also at room temperature, reducing the signal drift (FigureS12b, Supporting Information). A constant bias voltage of 1 V was

applied to the sensor, and the current change was simultaneouslyrecorded by a sourcemeter (2634B, Keithley) under gas exposure.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acsami.9b03296.

TEM images and UV−visible absorption spectrum ofCNTs and WS2; SEM images of pristine and CNT-coated cellulose paper; electrical characterization withrespect to the number of nanomaterial coatings;improvement of the sensing response by WS2 function-alization; comparison of performance parameters withthose of other nanomaterial-based flexible NO2 sensors;preparation and sensing responses of CNT−MoS2-coated cellulose paper; temperature coefficient ofresistance; and room-temperature NO2 sensing (PDF)

■ AUTHOR INFORMATIONCorresponding Author*E-mail: jwc@yu.ac.kr.ORCIDJungwook Choi: 0000-0001-5916-9714NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis research was supported by the Basic Science ResearchProgram through the National Research Foundation of Korea(NRF) , funded by the Min i s t ry o f Educa t ion(2016R1D1A1B03932028) and by the Ministry of Science,ICT and Future Planning (2017R1A4A1015581 and2019R1C1C1007840).

■ REFERENCES(1) Wang, X.; Dong, L.; Zhang, H.; Yu, R.; Pan, C.; Wang, Z. L.Recent Progress in Electronic Skin. Adv. Sci. 2015, 2, No. 1500169.(2) Hammock, M. L.; Chortos, A.; Tee, B. C.-K.; Tok, J. B.-H.; Bao,Z. 25th Anniversary Article: The Evolution of Electronic Skin (E-Skin): A Brief History, Design Considerations, and Recent Progress.Adv. Mater. 2013, 25, 5997−6038.(3) Bandodkar, A. J.; Jeerapan, I.; Wang, J. Wearable ChemicalSensors: Present Challenges and Future Prospects. ACS Sens. 2016, 1,464−482.(4) Siegel, A. C.; Phillips, S. T.; Dickey, M. D.; Lu, N.; Suo, Z.;Whitesides, G. M. Foldable Printed Circuit Boards on PaperSubstrates. Adv. Funct. Mater. 2010, 20, 28−35.(5) Tobjork, D.; Osterbacka, R. Paper Electronics. Adv. Mater. 2011,23, 1935−1961.(6) Wu, J.; Wu, Z.; Han, S.; Yang, B.-R.; Gui, X.; Liu, C.; Tao, K.;Miao, J.; Norford, L. K. Extremely Deformable, Transparent, andHigh-Performance Gas Sensor Based on Ionic Conductive Hydrogel.ACS Appl. Mater. Interfaces 2019, 11, 2364−2373.(7) Russo, A.; Ahn, B. Y.; Adams, J. J.; Duoss, E. B.; Bernhard, J. T.;Lewis, J. A. Pen-on-Paper Flexible Electronics. Adv. Mater. 2011, 23,3426−3430.(8) Nyholm, L.; Nystrom, G.; Mihranyan, A.; Strømme, M. TowardFlexible Polymer and Paper-Based Energy Storage Devices. Adv.Mater. 2011, 23, 3751−3769.(9) Ko, Y.; Kwon, M.; Bae, W. K.; Lee, B.; Lee, S. W.; Cho, J.Flexible Supercapacitor Electrodes Based on Real Metal-LikeCellulose Papers. Nat. Commun. 2017, 8, No. 536.(10) Zang, X.; Shen, C.; Chu, Y.; Li, B.; Wei, M.; Zhong, J.;Sanghadasa, M.; Lin, L. Laser-Induced Molybdenum Carbide-

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.9b03296ACS Appl. Mater. Interfaces 2019, 11, 19363−19371

19369

Graphene Composites for 3D Foldable Paper Electronics. Adv. Mater.2018, 30, No. 1800062.(11) Mahadeva, S. K.; Walus, K.; Stoeber, B. Paper as a Platform forSensing Applications and Other Devices: A Review. ACS Appl. Mater.Interfaces 2015, 7, 8345−8362.(12) Li, R.-Z.; Hu, A.; Zhang, T.; Oakes, K. D. Direct Writing onPaper of Foldable Capacitive Touch Pads with Silver Nanowire Inks.ACS Appl. Mater. Interfaces 2014, 6, 21721−21729.(13) Ahn, J.; Seo, J.-W.; Lee, T.-I.; Kwon, D.; Park, I.; Kim, T.-S.;Lee, J.-Y. Extremely Robust and Patternable Electrodes for Copy-Paper-Based Electronics. ACS Appl. Mater. Interfaces 2016, 8, 19031−19037.(14) Li, Y.; Samad, Y. A.; Taha, T.; Cai, G.; Fu, S.-Y.; Liao, K. HighlyFlexible Strain Sensor from Tissue Paper for Wearable Electronics.ACS Sustainable Chem. Eng. 2016, 4, 4288−4295.(15) Santhiago, M.; Correa, C. C.; Bernardes, J. S.; Pereira, M. P.;Oliveira, L. J. M.; Strauss, M.; Bufon, C. C. B. Flexible and FoldableFully-Printed Carbon Black Conductive Nanostructures on Paper forHigh-Performance Electronics, Electrochemical, and WearableDevices. ACS Appl. Mater. Interfaces 2017, 9, 24365−24372.(16) Saha, B.; Baek, S.; Lee, J. Highly Sensitive Bendable andFoldable Paper Sensors Based on Reduced Graphene Oxide. ACSAppl. Mater. Interfaces 2017, 9, 4658−4666.(17) Tao, L.-Q.; Zhang, K.-N.; Tian, H.; Liu, Y.; Wang, D.-Y.; Chen,Y.-Q.; Yang, Y.; Ren, T.-L. Graphene-Paper Pressure Sensor forDetecting Human Motions. ACS Nano 2017, 11, 8790−8795.(18) Yun, S.; Kim, J. Multi-Walled Carbon Nanotubes-CellulosePaper for a Chemical Vapor Sensor. Sens. Actuators, B 2010, 150,308−313.(19) Mirica, K. A.; Weis, J. G.; Schnorr, J. M.; Esser, B.; Swager, T.M. Mechanical Drawing of Gas Sensors on Paper. Angew. Chem., Int.Ed. 2012, 51, 10740−10745.(20) Han, J.-W.; Kim, B.; Li, J.; Meyyappan, M. A Carbon NanotubeBased Ammonia Sensor on Cellulose Paper. RSC Adv. 2014, 4, 549−553.(21) Han, J.-W.; Kim, B.; Li, J.; Meyyappan, M. Carbon NanotubeBased Humidity Sensor on Cellulose Paper. J. Phys. Chem. C 2012,116, 22094−22097.(22) Mirica, K. A.; Azzarelli, J. M.; Weis, J. G.; Schnorr, J. M.;Swager, T. M. Rapid Prototyping of Carbon-Based ChemiresistiveGas Sensors on Paper. Proc. Natl. Acad. Sci. U.S.A. 2013, 110, E3265−E3270.(23) Zhao, H.; Zhang, T.; Qi, R.; Dai, J.; Liu, S.; Fei, T. Drawn onPaper: A Reproducible Humidity Sensitive Device by Handwriting.ACS Appl. Mater. Interfaces 2017, 9, 28002−28009.(24) Yang, G.; Lee, C.; Kim, J.; Ren, F.; Pearton, S. J. FlexibleGraphene-Based Chemical Sensors on Paper Substrates. Phys. Chem.Chem. Phys. 2013, 15, 1798−1801.(25) Zhang, J.; Huang, L.; Lin, Y.; Chen, L.; Zeng, Z.; Shen, L.;Chen, Q.; Shi, W. Pencil-Trace on Printed Silver InterdigitatedElectrodes for Paper-Based NO2 Gas Sensors. Appl. Phys. Lett. 2015,106, No. 143101.(26) Jang, J.; Ha, J.; Cho, J. Fabrication of Water-DispersiblePolyaniline-Poly(4-styrenesulfonate) Nanoparticles For Inkjet-PrintedChemical-Sensor Applications. Adv. Mater. 2007, 19, 1772−1775.(27) Arena, A.; Donato, N.; Saitta, G.; Bonavita, A.; Rizzo, G.; Neri,G. Flexible Ethanol Sensors on Glossy Paper Substrates Operating atRoom Temperature. Sens. Actuators, B 2010, 145, 488−494.(28) Liu, H.; Li, M.; Voznyy, O.; Hu, L.; Fu, Q.; Zhou, D.; Xia, Z.;Sargent, E. H.; Tang, J. Physically Flexible, Rapid-Response GasSensor Based on Colloidal Quantum Dot Solids. Adv. Mater. 2014,26, 2718−2724.(29) Llobet, E. Gas Sensors Using Carbon Nanomaterials: A Review.Sens. Actuators, B 2013, 179, 32−45.(30) Kong, J.; Chapline, M. G.; Dai, H. Functionalized CarbonNanotubes for Molecular Hydrogen Sensors. Adv. Mater. 2001, 13,1384−1386.(31) Ko, K. Y.; Song, J.-G.; Kim, Y.; Choi, T.; Shin, S.; Lee, C. W.;Lee, K.; Koo, J.; Lee, H.; Kim, J.; Lee, T.; Park, J.; Kim, H.

Improvement of Gas-Sensing Performance of Large-Area TungstenDisulfide Nanosheets by Surface Functionalization. ACS Nano 2016,10, 9287−9296.(32) Cha, J.-H.; Choi, S.-J.; Yu, S.; Kim, I.-D. 2D WS2-EdgeFunctionalized Multi-Channel Carbon Nanofibers: Effect of WS2Edge-Abundant Structure on Room Temperature NO2 Sensing. J.Mater. Chem. A 2017, 5, 8725−8732.(33) Wang, Q. H.; Zadeh, K. K.; Kis, A.; Coleman, J. N.; Strano, M.S. Electronics and Optoelectronics of Two-Dimensional TransitionMetal Dichalcogenides. Nat. Nanotechnol. 2012, 7, 699−712.(34) Wu, W.; Wang, L.; Li, Y.; Zhang, F.; Lin, L.; Niu, S.; Chenet,D.; Zhang, X.; Hao, Y.; Heinz, T. F.; Hone, J.; Wang, Z. L.Piezoelectricity of Single-Atomic-Layer MoS2 for Energy Conversionand Piezotronics. Nature 2014, 514, 470−474.(35) Perkins, F. K.; Friedman, A. L.; Cobas, E.; Campbell, P. M.;Jernigan, G. G.; Jonker, B. T. Chemical Vapor Sensing withMonolayer MoS2. Nano Lett. 2013, 13, 668−673.(36) Lopez, N. P.; Elías, A. L.; Berkdemir, A.; Beltran, A. C.;Gutierrez, H. R.; Feng, S.; Lv, R.; Hayashi, T.; Urías, F. L.; Ghosh, S.;Muchharla, B.; Talapatra, S.; Terrones, H.; Terrones, M. PhotosensorDevice Based on Few-Layered WS2 Films. Adv. Funct. Mater. 2013,23, 5511−5517.(37) Choi, J.; Pyo, S.; Baek, D.-H.; Lee, J.-I.; Kim, J. Thickness-,Alignment- and Defect-Tunable Growth of Carbon Nanotube ArraysUsing Designed Mechanical Loads. Carbon 2014, 66, 126−133.(38) Zeng, H.; Liu, G.-B.; Dai, J.; Yan, Y.; Zhu, B.; He, R.; Xie, L.;Xu, S.; Chen, X.; Yao, W.; Cui, X. Optical Signature of SymmetryVariations and Spin-Valley Coupling in Atomically Thin TungstenDichalcogenides. Sci. Rep. 2013, 3, No. 1608.(39) Ago, H.; Kugler, T.; Cacialli, F.; Salaneck, W. R.; Shaffer, M. S.P.; Windle, A. H.; Friend, R. H. Work Functions and SurfaceFunctional Groups of Multiwall Carbon Nanotubes. J. Phys. Chem. B1999, 103, 8116−8121.(40) Choi, J.; Zhang, H.; Choi, J. H. Modulating OptoelectronicProperties of Two-Dimensional Transition Metal DichalcogenideSemiconductors by Photoinduced Charge Transfer. ACS Nano 2016,10, 1671−1680.(41) Lee, K.; Lee, J.-W.; Kim, S.-I.; Ju, B.-K. Single-Walled CarbonNanotube/Nafion Composites as Methanol Sensors. Carbon 2011,49, 787−792.(42) Choi, J.; Zhang, H.; Du, H.; Choi, J. H. Understanding SolventEffects on the Properties of Two-Dimensional Transition MetalDichalcogenides. ACS Appl. Mater. Interfaces 2016, 8, 8864−8869.(43) Ovchinnikov, D.; Allain, A.; Huang, Y.-S.; Dumcenco, D.; Kis,A. Electrical Transport Properties of Single-Layer WS2. ACS Nano2014, 8, 8174−8181.(44) Zhao, J.; Buldum, A.; Han, J.; Lu, J. P. Gas MoleculeAdsorption in Carbon Nanotubes and Nanotube Bundles. Nano-technology 2002, 13, 195−200.(45) Zhou, C.; Yang, W.; Zhu, H. Mechanism of Charge Transfer itsImpacts on Fermi-Level Pinning for Gas Molecules Adsorbed onMonolayer WS2. J. Chem. Phys. 2015, 142, No. 214704.(46) Langmuir, I. The Adsorption of Gases on Plane Surfaces ofGlass, Mica and Platinum. J. Am. Chem. Soc. 1918, 40, 1361−1403.(47) Wu, J.; Feng, S.; Wei, X.; Shen, J.; Lu, W.; Shi, H.; Tao, K.; Lu,S.; Sun, T.; Yu, L.; Du, C.; Miao, J.; Norford, L. K. Facile Synthesis of3D Graphene Flowers for Ultrasensitive and Highly Reversible GasSensing. Adv. Funct. Mater. 2016, 26, 7462−7469.(48) Wu, J.; Li, Z.; Tao, K.; Liu, C.; Khor, K. A.; Miao, K.; et al. 3DSuperhydrophobic Reduced Graphene Oxide for Activated NO2Sensing with Enhanced Immunity to Humidity. J. Mater. Chem. A2018, 6, 478−488.(49) Ou, J. Z.; Ge, W.; Carey, B.; Daeneke, T.; Rotbart, A.; Shan,W.; Wang, Y.; Fu, Z.; Chrimes, A. F.; Wlodarski, W.; Russo, S. P.; Li,Y. X.; Zadeh, K. K. Physisorption-Based Charge Transfer in Two-Dimensional SnS2 for Selective and Reversible NO2 Gas Sensing. ACSNano 2015, 9, 10313−10323.(50) Khan, H.; Zavabeti, A.; Wang, Y.; Harrison, C. J.; Carey, B. J.;Mohiuddin, M.; Chrimes, A. F.; Castro, I. A. D.; Zhang, B. Y.; Sabri,

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.9b03296ACS Appl. Mater. Interfaces 2019, 11, 19363−19371

19370

Y. M.; Bhargava, S. K.; Ou, J. Z.; Daeneke, T.; Russo, S. P.; Li, Y.;Zadeh, K. K. Quasi Physisorptive Two Dimensional Tungsten OxideNanosheets with Extraordinary Sensitivity and Selectivity to NO2.Nanoscale 2017, 9, 19162−19175.(51) Lee, S.; Reuveny, A.; Reeder, J.; Lee, S.; Jin, H.; Liu, Q.;Yokota, T.; Sekitani, T.; Isoyama, T.; Abe, Y.; Suo, Z.; Someya, T. ATransparent Bending-Insensitive Pressure Sensor. Nat. Nanotechnol.2016, 11, 472−478.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.9b03296ACS Appl. Mater. Interfaces 2019, 11, 19363−19371

19371