Abstract Carbon paste electrodes bulk-modified with Bi2O3were used for the determination of Cd(II) and Pb(II). Thebest composition was 1% (wt%) Bi2O3 in the paste. Themeasurements were made by differential pulse voltamme-try in the potential range from –1.2 V to –0.3 V. The peakpotential of the reoxidation of Cd is –0.85 V, and of Pb–0.60 V vs. SCE. The lowest concentration that could bedetermined was 5 µg L–1 of both metals (preconcentrationtime 240 s), the relative standard deviation was 3.5%–5.0%(four determinations). The correlation coefficient (r2) ofthe calibration curves was 0.9966 (for Cd) and 0.9971 (forPb). The Bi2O3-modified electrode could be used for theanalysis of drinking water, mineral water and urine.

Keywords Modified carbon paste electrode · Bi2O3 ·Stripping voltammetry · Cadmium · Lead

Introduction

Electrochemical stripping analysis is one of the best meth-ods for the detection of trace metals [1, 2]. A few metalscan be monitored simultaneously at concentrations down

to sub-ppb ranges. The carbon paste electrode (CPE) wasestablished by Adams 1959 and was initially used foranalysis by Jacobs in 1963 [3]; it can be used to determinecertain trace metals, and seems in this respect to be an al-ternative to mercury electrodes. To improve the detectionlimit of the plain electrode, it is modified by various com-pounds [4]. CPE is inexpensive, easy to use and it is notdifficult to regenerate the surface of it, in contrast withother solid electrodes [4, 5, 6].

Wang et al. used glassy-carbon and carbon fibre elec-trodes as a support for bismuth film[7, 8], and Krolicka etal. deposited bismuth film onto carbon paste surface [9],because Bi is an environmentally friendly element withlow toxicity. Such electrodes were applicable to strippinganalysis and were more sensitive for heavy metals thanthe same electrodes without films and a mercury filmelectrode.

Espinosa et al. studied the electrochemical propertiesof Bi2O3 mixing it with carbon powder in strongly alka-line solution [10]. They observed that a chemical reactionoccurs between bismuth oxide and OH– ions (Eq. 1), whilethe electrochemical reduction of BiO2

– is occurring at–0.8 V (Eq. 2):

Bi2O3(s) + 2OH− → 2BiO−2 + H2O (1)

BiO−2 + 2H2O + 3e → Bi(s) + 4OH− (2)

Another reduction can be observed at –1.0 V (vs. SCE),corresponding to:

Bi2O3(s) + 3H2O + 6e → 2Bi(s) + 6OH− (3)

There are no data in the literature regarding the use ofCPE modified with Bi compounds for the determinationof heavy metals.

Bulk-modified electrodes are simpler to prepare ascompared to film electrodes; additionally they are less ex-pensive than other solid electrodes.

The aim of the work presented here was to investigatethe possibilities of whether bismuth oxide-modified car-bon electrodes can substitute conventional mercury elec-trodes to some extent in environmental analysis.

Rasa Pauliukaite · Radovan Metelka · Ivan Švancara ·Agnieszka Królicka · Andrzej Bobrowski · Karel Vytřas ·Eugenijus Norkus · Kurt Kalcher

Carbon paste electrodes modified with Bi2O3 as sensors for the determination of Cd and Pb

Anal Bioanal Chem (2002) 374 :1155–1158DOI 10.1007/s00216-002-1569-3

Received: 21 January 2002 / Revised: 30 July 2002 / Accepted: 30 July 2002 / Published online: 24 October 2002

ORIGINAL PAPER

R. Pauliukaite · K. Kalcher (✉)Institute of Chemistry, Analytical Chemistry, Karl-Franzens University, 8010 Graz, Austriae-mail: kurt.kalcher@kfunigraz.ac.at

R. Metelka · I. Švancara · K. VytřasDepartment of Analytical Chemistry, Faculty of Chemical Technology, University of Pardubice, 53210 Pardubice, Czech Republic

A. Królicka · A. BobrowskiDepartment of Building Materials, Faculty of Materials Engineering and Ceramics, University of Mining and Metallurgy, 30–059 Krakow, Poland

R. Pauliukaite · E. NorkusInstitute of Chemistry, Department of Kinetics and Catalysis,2600 Vilnius, Lithuania

© Springer-Verlag 2002

Experimental

Apparatus

Differential pulse anodic stripping voltammetry was performedwith a polarographic analyzer (Model 264A EG&G Princeton Ap-plied Research, USA), equipped with an X–Y recorder (Siemens,Kompensograph C-1924). A carbon paste electrode (6 mm diame-ter) served as the working electrode, a saturated calomel electrode(SCE) and a platinum wire were used as the reference and counterelectrodes, respectively.

Reagents and chemicals

Water was distilled twice in a quartz still and finally purified witha cartridge system (Nanopure, Barnstead, USA). All solutions wereprepared with highly pure water. Stock solutions of cadmium andlead (1000 mg L–1) were obtained from Merck (Germany) and di-luted as required. An acetate buffer solution (0.1 M, pH 4.5) wasused as the supporting electrolyte [7].

Working electrodes

The carbon paste was made from carbon powder (RW-B, Rings-dorff-Werke Ltd., Germany) and liquid paraffin (Uvasol, Merck,Germany) by mixing 400 mg carbon powder with 90 µL paraffin[6]. After thorough mixing, 1% Bi2O3 (wt%) was added to thepaste. A freshly modified CPE was prepared every day.

Procedure

Stripping voltammetric measurements were performed withoutdeaeration of solutions. The potential of preconcentration of Cdand Pb was usually –1.2 V, with a usual preconcentration time of240 s. The equilibration time was 15 s, after which the voltammo-gram in differential pulse mode was recorded with an anodic scan(sampling time 0.2 s per data, pulse height 25 mV). The scan wasterminated at –0.3 V. After each measurement the electrode sur-face was renewed.

Samples

Tap water, naturally carbonated mineral water (“Vöslauer”, Aus-tria) and urine were used as samples. The tap water was collectedin the laboratory and pretreated before analysis; it was evaporatedwith concentrated HNO3 to dryness to decompose dissolved or-ganic matter (due to interference of organic materials present indrinking water on analysis), and redissolved in one-tenth of the ini-tial volume of the sample. The mineral water was treated in thesame way, but redissolved in the same volume, as had the initialsample. Human urine was first kept in a refrigerator for 10 h, thenfiltered and finally was mineralized with concentrated HNO3 in thesame way as mineral water.

Results and discussions

The unmodified CPE can be used for the determination ofCd(II) and Pb(II) in the potential range from –1.2 to 0.0 Vin a supporting electrolyte of acetate buffer (0.1 M, pH 4.5).The corresponding stripping volammograms are shown inFig.1 (curves a, b). The reoxidation peaks of Cd and Pbappear at potentials of –0.85 V and –0.65 V, respectively.Reduction of oxygen containing carbon groups proceedsat negative potentials (-1.0 V) [6]. The responses at un-

modified CPEs are very low, therefore trace concentra-tions of the analytes cannot be determined.

To improve the limit of quantification of CPEs we triedto deposit a Bi-film on the CPE surface, but the effect ofsuch films was low [7]. The electrochemical activity ofBi-films depends on the structure of the electrode surface[7, 11], which is not ideal with CPEs for external platingof a film. More effective is bulk modification of the elec-trode by Bi2O3 (Fig.1, curve c). In this case the potentialof the reoxidation of Cd is the same and the reoxidation ofPb is shifted to a somewhat less negative potential at–0.58 V. The lower scan limit was set to –0.3 V, becausethe reoxidation of Bi occurs beyond this potential. Thebest response of the modified CPE was obtained by mod-ification with 1% of Bi2O3 (wt%). Lower and higher con-centrations yielded decreased signals. Obviously the physi-cochemical properties of the electrode material are changedsignificantly by the addition of the modifier yielding max-imum response with concentration of 1% Bi2O3 (wt%) inthe paste. Attempts to determine Zn(II) and Tl(I) in thesame solution were not very successful, as this electrodeis not sensitive enough for Zn and the reoxidation of Cdand Tl occurs at the same potential.

The electrochemical reduction of Bi2O3 (Eq. 2) occursat –1.0 V in 0.1 M acetate buffer (pH 4.5) (Fig.1, curve c).This signal is shifted to more negative potentials com-pared with data from Espinosa et al. [10], because of theslightly acidic solution. The reoxidation peak of Cd(II), inthe presence and absence of Pb(II) in the solution, is foundat the same potential with a modified and an unmodifiedCPE. Some data in literature indicate that Cd forms in-termetallic compounds with Bi [12, 13]; with higher con-centrations of Pb, cadmium, which is also present, willhave diminished possibilities for adsorption [13]. Alka-line solutions as well as more acidic solutions yieldedsignals and noise with the blank solution already and,

1156

Fig.1 Differential pulse anodic stripping voltammograms of Cdand Pb with an unmodified CPE in absence of metals (a); with anunmodified CPE and 50 µg L–1 of each metal (b), and with a CPEbulk-modified with Bi2O3 and 50 µg L–1 of each metal (c). Sup-porting electrolyte: acetate buffer (0.1 mol L–1, pH 4.5). Precon-centration time 240 s at –1.20 V, scan rate 5 mV s–1, pulse high25 mV

therefore, could not be used for the accumulation of Cdand Pb.

For lead, the reoxidation currents are the same, inde-pendent of the presence or absence of cadmium. To somesmall extent, Pb can also be connected to the formation ofa mixed compound with bismuth oxide [14, 15], becausethere are concentration dependent potential-shifts notice-able: the peak potential of 5 µg L–1 Pb is at –0.65 V, whereasit is shifted to –0.6 V with 20 µg L–1.

Using a preconcentration time of 120 s, the calibrationcurve of Cd(II) is not linear within the range from 10 to100 µg L–1, and there are deviations of peak currents(8%–10%) with the same concentration. Probably it takessome time until a stable Bi0-film is formed from Bi2O3 inthe bulk of the carbon paste. This disadvantage can beeliminated by increasing the preconcentration time forboth metals to 240 s. The dependence of the signal on theaccumulation time was linear up to 10 min. Both calibra-tions curves are then linear and well reproducible (Figs. 2shows the calibration curve for Pb). The relative standarddeviation is 3.5%–5.0% for Cd and Pb (20 and 50 µg L–1,four determinations). The correlation coefficient (r2) ofthe calibration curve is 0.9966 (for Cd) and 0.9971 (forPb) and the regression equations are: y=1.3149x and

y=1.1151x (where y is the reoxidation current in µA and xis the concentration of Cd or Pb in µg L–1) for Cd and Pb,respectively.

The reoxidation peak of Cd was lower than the one ofPb when the preconcentration time was 240 s. This isprobably due to the adsorption of Pb on the modified CPEor stronger binding to bismuth oxide [14]. The peak of Cdreoxidation is higher than the response of lead only in thelow concentration range, when Pb covers little part of thesurface of the electrode, and the signals of both metals areequal when their concentration is 20 µg L–1.

The modified electrode was tested for the determina-tion of Cd and Pb in tap water, mineral water, and urine.Drinking water was mineralized with concentrated HNO3,because some trace of organic material present in drinkingwater interferes with the analysis; the dry mass was redis-solved in a tenth of the original volume. Lead could be de-tected in drinking water from the pipe, but Cd was belowthe limit of detection (Table 1).

Interferences from other metal ions were investigated.It turned out that mainly heavy metals, which interferealso with mercury-film electrodes, showed an interferenceon the signal of Cd and Pb with the Bi2O3-modified car-bon paste electrode. Thus, Tl overlapped with the signalof Cd at higher concentrations. Similarly, indium gave asignal at the same position as cadmium; therefore bothmetals have to be absent in the sample solution.

An Austrian mineral water was also analyzed. It wasdecarbonated and mineralized with HNO3. Both metals, Cdand Pb, were determined by the standard addition method(Table 1).

Urine was mineralized with concentrated HNO3. Leadcould be determined without problems, but the signal ofan unknown compound interfered with Cd (Table 1).

All results were compared with reference determina-tions carried out using ICP-MS. The results of the watersamples differed by about +3% (higher with the electro-chemical method), whereas the result of urine had a devi-ation of +5.5%.

Concerning practical applications of voltammetric sen-sors based on modifications with bismuth oxide, theirlimit of detection of heavy metals under investigation issomehow comparable to the one achievable with graphitefurnace atomic absorption spectrometry (GFAAS). Themain advantage of the sensors over the optical method istheir mobility with respect to direct field applications as

1157

Fig.2 Calibration curves Pb(II) with a CPE bulk-modified withBi2O3. Supporting electrolyte acetate buffer (0.1 mol L–1, pH 4.5).Deposition time 240 s at –1.20 V, scan rate 5 mV s–1, pulse high25 mV

Table 1 Results of analysis of drinking water and urine and norm values for Cd and Pb. SPCE modified by Bi2O3 (1% wt%). Support-ing electrolyte: acetate buffer (0.1 mol l–1, pH 4.5). Preconcentration time 240 s, pulse high 25 mV, scan rate 5 mV s–1

Sample CPEs results (µg L–1) CPEs results (standard addition) (µg L–1) Norms (µg L–1)

Cd Pb Cd Pb Cd Pb

Drinking water – 19.3±0.6a – 1.65±0.05 Up to 5c Up to 15c

Mineral water 1.82±0.05 1.88±0.06Urine –b 117.7±3.5 Up to 2.6d 80–120d

aResults obtained in the one-tenth volume of original sample.bDue to interference it was not possible to determine Cd in urine.

cBy Environmental Protection Agency (EPA).dFrom [15].

single or multi-shot sensor in combination with hand-helddevices. A small drawback concerning the determinationof the heavy metals Cd and Pb seems the need of a miner-alization procedure, but also this can be easily realized inthe field by means of simple gas burners. Work is inprogress to investigate possibilities to replace chemicalmineralization by simpler chemical, photochemical, or byelectrochemical oxidation. Therefore, the sensors devel-oped here are very promising for on-site applications,maybe even just in a simple digital way to provide infor-mation whether a sample should be analyzed more closelyin the laboratory or not.

Conclusions

Carbon paste electrodes bulk modified by Bi2O3 can beused for monitoring Cd(II) and Pb(II) simultaneouslydown to a concentration range 5 µg L–1.

Such electrodes are very promising candidates for sub-stitutes of sensors based on mercury. They are easily andquickly prepared using inexpensive and nontoxic materi-als. Above this, they offer the advantage that they can beeven modified with additional additives, thus allowing thepreparation of the more complex sensors and biosensors.This will be the subject of the future work.

Acknowledgements The authors express their thanks to the fol-lowing funds for financial support: Austrian Academic ExchangeService (OEAD, North-South Dialogue Program), CEEPUS pro-gram (PL-110) and Ministry of Education Youth and Sport of theCzech Republic. Great thanks to Prof. Jan Mocak (Slovak Techni-cal University, Bratislava, Slovakia) for discussions.

References

1.Wang J (1985) Stripping analysis. VCH Publishers, DeerfieldBeach

2.Achtenberg EP, Braungardt C (1999) Anal Chim Acta 400:3813. Jacobs ES (1963) Anal Chem 35:21124.Kalcher K, Kauffmann JM, Wang J, Švancara I, Vytras K,

Neuhold C, Yang Z (1995) Electroanalysis 7:55.Švancara I, Zima J, Schachl K (1998) Scient Papers Univ Par-

dubice A 4:496.Švancara I, Schachl K (1999) Chem Listy 93:4907.Wang J, Lu J, Hocevar SB, Farias PAM, Ogorevc B (2000)

Anal Chem 72:32188.Hocevar S, Ogorevc B, Wang J (2000) 7th Young Investiga-

tors’ Seminar on Analytical Chemistry, Book of Abstracts,Graz, Austria

9.Królicka A, Pauliukaité R, Švancara I, Metelka R, BobrowskiA, Norkus E, Kalcher K, Vytřas K (2002) Electrochem Com-mun 4:193

10. Espinosa AM, San José MT, Tascón ML, Vázquez MD, SánchezBatanero P (1991) Electrochim Acta 36:1561

11.Besse F, Boulanger C, Lecuire JM (2000) J Appl Electrochem30:385

12.Cotton IA, Wilkinson G (1980) Advanced inorganic chemistry,4th edn. Wiley–Interscience, New York

13.Greenwood N, Earnshaw A (1988) Chemie der Elemente. VCHVerlagsgeselschaft, Weinheim, Germany

14.Espinosa AM, Tascón ML, Encinas P, Vázquez MD, Sánchez-Batanero P (1995) Electrochim Acta 40:1623

15.Vivier V, Régis A, Sagon G, Nedelec JY, Yu LT, Cachet-Vivier C (2001) Electrochim Acta 46:907

1158

This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 69.91.157.17

This content was downloaded on 02/04/2015 at 18:13

Please note that terms and conditions apply.

Surface micromachining technology applied to the fabrication of a FET pressure sensor

View the table of contents for this issue, or go to the journal homepage for more

1996 J. Micromech. Microeng. 6 80

(http://iopscience.iop.org/0960-1317/6/1/018)

Home Search Collections Journals About Contact us My IOPscience

J. Micromech. Microeng. 6 (1996) 80–83. Printed in the UK

Surface micromachining technologyapplied to the fabrication of a FETpressure sensor

L Svensson †, J A Plaza , M A Benitez, J Esteve and ELora-Tamayo

Centro Nacional de Microelectronica-CNM, CSIC, Campus UAB, 08193 Bellaterra,Spain

Received 7 December 1995, accepted for publication 28 December 1995

Abstract. This paper presents a novel surface-micromachined pressure sensorbased in a FET device. The diaphragm acts as the gate of the transistor and thegate–source voltage varies in a nearly linear form with pressure when keepingconstant the current along the channel. Both theoretical characteristics and thetechnological process are analysed. Finally the advantages over existingmicrosensors are examined.

1. Introduction

Surface micromachining is one of the most promisingtechnologies employed in microsensor fabrication and it isalready giving fruits as commercial products [1, 2]. Somemechanical structures have been fabricated that are sensitiveto a wide range of physical parameters. This paper dealswith a static absolute FET pressure sensor. It consistsof a sealed vacuum cavity with a 2µm thin deflectablepolysilicon diaphragm. Polysilicon was chosen on accountof its good mechanical and electrical qualities. In order todetect the applied static pressure the MIS (metal–insulator–semiconductor) system capacitance principle is used, butwith the difference that now there is an insulator withvariable thickness. The pressure is applied to the upperelectrode.

The pressure sensor is in essence a FET transistor witha dielectric of variable thickness where the membrane is thedevice gate and the two diffusions under the dielectric arethe drain and the source regions. The dielectric is formedby a vacuum cavity and two layers, one of silicon nitride,and another of silicon dioxide below it.

Although some FET pressure sensors have beenrealized recently [3, 4], they have a complex process anda poor sensitivity. Also surface-micromachined capacitivepressure sensors have been made [5], but the FET structureallows us to fabricate sensors smaller than the capacitiveones, due to the large area needed in the latter to achievea capacitance of the order of a few picofarads.

† e-mail: svensson@cnm.es.

2. Description of the technological process

For the sensor fabrication the starting material is p-doped〈100〉-oriented silicon wafers. In figure 1 the fabricationprocess is shown. First a boron field implantation is carriedout, followed by source and drain implantations. It is alsonecessary to adjust the threshold voltage in the channelregion. The next step is to form two layers, one of thermalsilicon oxide and another of silicon nitride. Following, thecontacts are opened in two steps. Then the sacrificial layeris patterned, leaving oxide in the future cavity below themembrane. In the next step a thin oxide layer is depositedand patterned to define the membrane anchoring areas.After that, a polysilicon layer is deposited, doped withphosphorus and annealed, to achieve a low tensile stressand a minimum stress gradient [6]. Then, the membraneis patterned by a RIE process. Afterwards, the sacrificialoxide is etched with HF. Then metal is deposited and theattack windows are sealed. Finally, the metal lines arepatterned.

3. Physical model and starting hypothesis

The simple physical model of a MISFET is used. Theequation relating membrane deflexionw to pressureP andradial distancer from center is [7]

w(r) = P(a2 − r2)2

64D

wherea is the membrane radius andD the flexural rigidity.For a high sensitivity, the channel must be located underthe maximum deflexion point or near it. Regrettably, themembrane bending is not constant along a radial direction,so there will be variation of the distance between the

0960-1317/96/010080+04$19.50 c© 1996 IOP Publishing Ltd

Surface-micromachined FET pressure sensor

Figure 1. Basic steps of the fabrication process.

membrane and the channel along the channel. One wayto avoid this is to take into account the radial symmetry ofthe diaphragm; thus a ring channel design was chosen. Theapproximation of flat bending along the channel length wastaken.

The expression of the threshold voltage is:

Vt(P ) = 8ms + 28f n + Qb − Qss

Cs(P )

where 8ms is the metal-semiconductor work function

Figure 2. Plot of Vgs(P) for different Ids .

difference,8f n the Fermi level potential difference betweenthe inversion region and the bulk,Qb the charge densityin the depletion region,Qss the charge density at the Si–SiO2 interface andCs the capacitance per unit area . Theexpression ofCs as a pressure is

Cs(P ) =(C−1

ox + C−1nit + C−1

vac(P ))−1

where

Cvac(P ) = ε0

h − kP.

This is the serial combination of the capacitance perunit area of the silicon oxideCox , the silicon nitrideCnit

and the variable gap in the cavityCvac, whereh is the gapdistance atP = 0 andk the constant relating the deflexionw to pressureP . The threshold voltage shows a linearrelation with pressureP .

The expression of the gate–source voltage in saturationmode and keeping constantIds is

Vgs(P ) = Vi(P ) +√

2LIds

µWCs(P )

whereµ is the surface mobility,W the channel width andL the channel length. This expression is approximatelylinear for a low drain–source current and a theoretical

81

L Svensson et al

Figure 3. Photograph of the chip containing several pressure sensors.

Figure 4. A SEM photograph of the polysilicon membrane.

sensitivity of about 13 V bar−1 is achieved for the sensorwith a membrane radius of 86µm. For the fabricatedcircular-channel sensorsL = 5 µm and W = 45 µm.Figure 2 shows the theoretical gate voltage versus pressurefor different drain currents (radius 52µm).

In figure 3 a photograph of the fabricated chip isshown, containing several pressure sensors with differentdimensions, each one works at a specified pressure range.The membrane diameters are 40, 104, 148, 154 and172µm. The two devices located at the upper right cornerare reference devices. The chip size is 1920× 1920 µm2

.In figure 4 a SEM photograph of a pressure sensor (radius52 µm) is shown.

Figure 5. Ids –Vgs(P) experimental measurement.

4. Conclusions

A FET pressure sensor has been fabricated based on amodified surface-micromachined technique. Sensors ofdifferent membrane sizes have been measured, with circularor linear channels. ElectricalIds–Vgs measurements weremade, with the linear channel 86µm radius sensorgiving the best performance (see figure 5). It worksin a two-mode operation: free bending (1–2 bar), andground touching (2–4 bar), and a zero-sensitivity point isobserved atVgs = 8.2 V. It has a high sensitivity withpressure (0.1 mA bar−1). Surface micromachining hasadvantages over other technologies: an accurate structurallayer thickness control, and the possibility of making verysmall devices.

82

Surface-micromachined FET pressure sensor

References

[1] Core T A, Tsang W K and Sherman S J 1993Solid StateTechnol.Oct 39–47

[2] Kuehnel W and Sherman S 1994Sensors ActuatorsA 457–16

[3] Suminto J T and Ko W H 1990Sensors ActuatorsA21–23126–32

[4] Voorthuyzen J A and Bergveld P 1988Sensors Actuators14349–60

[5] Dudaicevs H, Kandler M, Manoli Y, Mokwa W and SpiegelE 1994Sensors ActuatorsA 43 157–63

[6] Benitez M A, Esteve J, Benrakkad M S, Morante J R,Samitier J and Schweitz J A 1995Transducers ’95 (June1995, Stockholm, Sweden)

[7] Timoshenko S 1959Theory of Plates and Shells

83

Skin-like pressure and strain sensors based ontransparent elastic films of carbon nanotubesDarren J. Lipomi1†, Michael Vosgueritchian1†, Benjamin C-K. Tee2†, Sondra L. Hellstrom3,

Jennifer A. Lee1, Courtney H. Fox1 and Zhenan Bao1*

Transparent, elastic conductors are essential components ofelectronic and optoelectronic devices that facilitate humaninteraction and biofeedback, such as interactive electronics1,implantable medical devices2 and robotic systems withhuman-like sensing capabilities3. The availability of conductingthin films with these properties could lead to the developmentof skin-like sensors4 that stretch reversibly, sense pressure(not just touch), bend into hairpin turns, integrate with collap-sible, stretchable and mechanically robust displays5 and solarcells6, and also wrap around non-planar and biological7–9 sur-faces such as skin10 and organs11, without wrinkling. Wereport transparent, conducting spray-deposited films ofsingle-walled carbon nanotubes that can be rendered stretch-able by applying strain along each axis, and then releasingthis strain. This process produces spring-like structures in thenanotubes that accommodate strains of up to 150% anddemonstrate conductivities as high as 2,200 S cm21 in thestretched state. We also use the nanotube films as electrodesin arrays of transparent, stretchable capacitors, which behaveas pressure and strain sensors.

Metallic films on elastomeric substrates can accommodate strainby means of controlled fracture12 or buckling13, but they are gener-ally opaque. Conductive polymers can be buckled to form stretch-able transparent electrodes6,14, but topographic buckles may beincompatible with devices that require planar interfaces. Films ofcarbon nanotubes15 and graphene16 are candidates for stretchable,transparent electrodes because (1) the long mean-free path of elec-trons in defect-free films produces high conductivity, withoutdecreasing the transparency17, and (2) networks of nanotubes andgraphene sheets permit some elasticity without destroying the con-tiguity of the film. One collaboration16,18 has produced graphenesheets with values of transparency T and sheet resistance Rsapproaching those of tin-doped indium oxide (ITO), but the resist-ance increased by an order of magnitude when strained by 30%(ref. 16). Others have demonstrated uniaxial stretchability inhighly aligned films of nanotubes pulled from vertical forests19.There have also been reports of films of nanotubes that can bestretched by up to 100% along the axis of aligned nanotubeswithout a significant change in resistance15. Randomly depositedfilms of nanotubes have been stretched up to 700%, but the resis-tance increased by an order of magnitude following the applicationof ≤50% strain20. Recently, researchers reported a transparentnanotube film embedded in an elastomer for stretchable organiclight-emitting devices; the resistance of the most conductive film(50 V sq21 at T¼ 63%) increased by 100% at 50% strain21.Stretchable, opaque networks of conductive particles demonstratedthus far include a nanotube–fluoroelastomer composite with

conductivity of 9.7 S cm21 at 118% strain5, a nanotube–silvercomposite material with 20 S cm21 at 140% (ref. 22) and bucklednanotube films with 900 S cm21 at 40% (ref. 23). Combining highconductivity (s . 100 S cm21) and transparency (.80%) at highstrain (1≥ 150%) remains a challenge.

We produced conductive, transparent, stretchable nanotubefilms by spray-coating (nanotube length¼ 2–3 mm)24,25 directlyonto a substrate of poly(dimethylsiloxane) (PDMS, activated byexposure to ultraviolet/O3) from a solution in N-methylpyrroli-done. We obtained the best values of Rs and s by spin-coating asolution of charge-transfer dopant (tetrafluorotetracyanoquinodi-methane (F4TCNQ) in chloroform) over the films26. Doped andundoped films exhibited similar electromechanical behaviour. Weobtained values of Rs¼ 328 V sq21 and T¼ 79%, and maximumvalues of s¼ 1,100 S cm21 for a 100 nm film with T¼ 68%, at0% strain. (See Supplementary Fig. S1 for the measurement offilm thickness.)

Figure 1a presents the evolution of the change in resistance(DR/R0) as a function of strain for seven stages of applied strainand relaxation: 0 � 50% � 0% � 100% � 0% � 150% � 0%� 200%. With the first application of 50% strain, R increased by0.71. We attribute this increase to irrecoverable loss of junctionsbetween nanotubes. When we returned the film to 0% strain,DR/R0 decreased to 0.64 (as opposed to 0, its original value).Following the application of 100% strain, the resistance retraceditself until it reached 50% (the previous maximum strain), afterwhich the slope of DR/R0 increased. We observed similar behaviourwhen we relaxed the film on reaching 1¼ 100% and 150%. At 1≈170%, the resistance of the film increased irreversibly by severalorders of magnitude. We estimate a lower limit on conductivity at150% strain of 2,200 S cm21 (see Supplementary Information fora discussion of conductivity under strain).

The effect of strain history on resistance implies that these nano-tube films can be ‘programmed’ by the first cycle of strain andrelease, to be reversibly stretchable within the range defined by thefirst strain. Figure 1b shows four cycles of strain from 0 to 50%, inwhich the resistance increases reversibly by ≤10% because thefilm was previously strained to a maximum of 50%. Measurementof R over the course of 12,500 cycles of stretching produced theplot shown in Fig. 1c. Over the course of this experiment, the resis-tance had decreased by 22% at the 1,500th cycle, and then increasedlinearly. We observed the same minimum in resistance at �1,000cycles of stretching in three similar experiments. We attribute theminimum in resistance to a period in which the nanotubebundles adopted their optimum morphology. Subsequent cycles ofstretching possibly decreased the number of conductive junctionsbetween bundles.

1Department of Chemical Engineering, Stanford University, Stanford, California 94305, USA, 2Department of Electrical Engineering, Stanford University,Stanford, California 94305, USA, 3Department of Applied Physics, Stanford University, Stanford, California 94305, USA; †These authors contributed equallyto this work. *e-mail: zbao@stanford.edu

LETTERSPUBLISHED ONLINE: 23 OCTOBER 2011 | DOI: 10.1038/NNANO.2011.184

NATURE NANOTECHNOLOGY | VOL 6 | DECEMBER 2011 | www.nature.com/naturenanotechnology788

© 2011 Macmillan Publishers Limited. All rights reserved.

We examined the morphology of the nanotube films usingatomic force microscopy (AFM) to understand why the resistanceof the film was a function of its strain history. Figure 2 shows aseries of schematic diagrams depicting the change in morphologyof nanotubes on a PDMS substrate with strain, as well as corre-sponding AFM images. The ‘as-deposited’ film (Fig. 2a) exhibitedbundles (diameter, 10–20 nm) of nanotubes with isotropic orien-tations. Activation of the surface with ultraviolet/O3 before depo-sition was critical, because the nanotube bundles formed large,sparse aggregates non-uniformly on hydrophobic substrates.Figure 2b shows an AFM image of the nanotube film understrain. The application of strain exerted tensile stress on bundleswith components oriented with the axis of strain and alignedthem to it (Fig. 2b, dashed box). Compressive stress (due to thePoisson effect) on bundles oriented perpendicular to the axis ofstrain caused them to buckle in plane into waves (Fig. 2b, solidbox). After stretching the film for the first time, relaxation to 0%strain produced waves in the bundles that had been aligned bystretching (Fig. 2c). The amplitude of waves increased with theinitial strain (Supplementary Fig. S2). Other researchers have

observed a different, although similar, phenomenon, with the buck-ling of individual nanotubes (as opposed to bundles) on the surfaceof PDMS under small compressive strains of 5% (as opposed to150%), with buckling amplitudes ,10 nm and perpendicular (asopposed to parallel) to the plane of the substrate27.

The stretch-induced change in the morphology of the uniaxiallystretched films produced unequal conductivities along the stretchedand unstretched axes. When both axes were stretched and released,all nanotube bundles exhibited buckling, but the orientations wererandom (Fig. 2d), and the resistance was the same along the stretchedand unstretched axes. Biaxially stretched films were reversibly stretch-able in any direction (see Supplementary Information and Fig. S3).

The technological goal is to integrate these stretchable, trans-parent conductors into interactive optoelectronic devices andsensors for biofeedback. We therefore fabricated transparent andstretchable parallel-plate capacitors that could manifest changes inpressure and strain as changes in capacitance (Fig. 3a). Thedevices comprised two strips of PDMS bearing stretchable nanotubefilms, which we laminated together, face-to-face, with Ecoflexsilicone elastomer. Ecoflex (Shore hardness 00-10) is more easily

0

1

2

3

4

5

0 50 100 150 200

No. of stretches

Stre

tch

to 5

0% Relax to 0%

ε (%)

t (s)

Rmin at 1,500th cycleR (kΩ

)ΔR

/R0

ΔR/R

0

a

b

c

−0.02

0

0.02

0.04

0.06

0.08

0.1

0 10 20 30 40 50

0

20

40

60

80

100

120

1 10 100 1,000 10,000

Figure 1 | Effects of applied strain on films of spray-coated carbon

nanotubes on PDMS substrates. a, Change in resistance DR/R0 versus

strain 1 for a nanotube film on a PDMS substrate. When the film is strained

(arrow, bottom left), DR/R0 increases, and remains constant as the strain is

released. When the strain is increased again, DR/R0 remains constant, and

then increases when 1 exceeds the value at which the strain was released

before. This sequence is repeated up to DR/R0 ≈ 5 and 1≈ 150%. b, DR/R0

versus time in response to four cycles of stretching from 0 to 50%.

c, Resistance versus number of stretches (on a log scale) over 12,500

cycles of stretching to 25%.

Compression

Carbon nanotube bundles: random orientation

Stretch-aligned

Buckled along stretched axis

Isotropic

Aligned with axis of strain

Buckled along axis of strain

Buckled, isotropic

b Stretched

a As-deposited

c Relaxed

d Biaxially stretched: relaxed

Buckled in all directions

PDMS

Tension

Figure 2 | Evolution of morphology of films of carbon nanotubes with

stretching. Schematics (left) and corresponding AFM phase images (right)

of nanotube films as deposited (a), under strain (b), stretched and released

along one axis (c), and stretched and released along two axes (d).

The bundles are considerably longer than the individual nanotubes

within them. Dashed and solid white boxes highlight the bundles of

nanotubes buckled along the horizontal and vertical axes, respectively.

Scale bars, 600 nm.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.184 LETTERS

NATURE NANOTECHNOLOGY | VOL 6 | DECEMBER 2011 | www.nature.com/naturenanotechnology 789

© 2011 Macmillan Publishers Limited. All rights reserved.

deformed than PDMS (Shore hardness A-48)28. The capacitance ofa parallel-plate capacitor is proportional to 1/d, where d is thespacing between plates. Application of pressure (Fig. 3a (left), b,d)and tensile strain (Fig. 3a (right), c, e) both resulted in a shorteneddistance between the electrodes (d′). Capacitance C is linearlydependent on pressure to 1 MPa (Fig. 3b) and strain to 50%(Fig. 3c) over the ranges tested. The smallest change in capacitancedistinguishable from noise was �50 kPa. The figure of merit of con-ventional strain gauges is the gauge factor, (DR/R0)/1 (ref. 29). Wecan also define a capacitive gauge factor, (DC/C0)/1, which is theslope of the linear fit in Fig. 3c (see Supplementary Informationfor discussion); in this case we determined (DC/C0)/1 to be0.004. Figure 3d shows capacitance versus time for four cycles ofapplied pressure using an electrically insulating tip to apply theload. Figure 3e shows a similar plot of capacitance versus timeover four cycles of stretching to 30%. The timescale over whichthe pixels recovered was smaller than that over which our instru-ments could load and unload the sample, ≤125 ms.

We next fabricated a grid of capacitors to produce a device thathad spatial resolution (Fig. 4). We began by depositing nanotubelines through a PDMS membrane that contained apertures (step 1)30.Applying strain rendered the film reversibly stretchable (step 2).We positioned a second patterned substrate orthogonal to the first(step 3), placed liquid eutectic gallium-indium (EGaIn)28,31,32 andcopper wires at the ends of the nanotube lines, and laminated thesubstrates together with Ecoflex (step 4).

We formed patterns of nanotube films with linewidths of 0.6–2 mm and pitches of 2–4 mm. We generated arrays of 4–64 capaci-tors (‘pixels’) with areas of 0.4–4 mm2 and pitches of 2–4 mm. Thethickness of the Ecoflex layer was �300 mm. Figure 5a,b shows thelargest array we fabricated: an 8 × 8 array of nanotube lines, withwidth and spacing of 2 mm. The average capacitance of each pixelwas 13.3+1.4 pF (N¼ 64). Transparency of the nanotube linesvaried across the substrate from 88 to 95%.

In principle, changes in capacitance due to strain could be dis-tinguished from those due to pressure. Tensile straining wouldaffect pixels along the axis of strain; pressure would affect thepixels in the immediate vicinity of the load. We found that thecrosstalk between adjacent pixels in the 64-pixel device was low,and the change in capacitance registered by the pixel on whichpressure was applied was five times higher than the average ofthat registered by the four adjacent pixels (Fig. 5c). This obser-vation highlights the advantages of using stretchable materialsfor pressure sensors, for which the greatest compression occursat the site of the load. The crosstalk for devices fabricated on a

d

d’ d’

a

Nanotube films

PDMS

PDMS

Unstrained profile

Ecoflex

P (MPa)

ΔC/C

0

ΔC/C

0

b

0

0.05

0.1

0.15

0.2

0 10 20 30 400

0.05

0.1

0.15

0.2

t (s)0 10 20 30 40

t (s)

ΔC/C

0

ΔC/C

0

Stre

tch

30%

Relax

App

ly 1

MPa U

nload

d e

ε (%)

c

y = 0.23xR2 = 0.994

y = 0.0041xR2 = 0.969

00.05

0.10.150.2

0.25

0 0.2 0.4 0.6 0.8 10

0.050.1

0.150.2

0.25

0 10 20 30 40 50

Figure 3 | Use of stretchable nanotube films in compressible capacitors that can sense pressure and strain. a, Schematic showing a stretchable capacitor

with transparent electrode (top), and the same capacitor after being placed under pressure (left) and being stretched (right). b,c, Change in capacitance

DC/C0 versus pressure P (b) and strain 1 (e). d,e, DC/C0 versus time t over four cycles of applied pressure (d) and stretching (e).

Step 1

Step 2

Step 3

Step 4

PDMS Nanotubes

0.6−2 mm

0.2−0.5 mm

Wires

Ecoflex(~300 μm)

EGaIn

Figure 4 | Summary of processes used to fabricate arrays of transparent,

compressible, capacitive sensors. Spray-coating through a stencil mask

produces lines of randomly oriented nanotubes (step 1). A one-time

application of strain and release produces waves in the direction of strain

(step 2). A second patterned substrate is positioned (face to face) over the

first (step 3). The two substrates are bonded together using Ecoflex silicone

elastomer, which, when cured, serves as a compressible dielectric layer

(step 4). Drops of a liquid metal, EGaIn, make conformal contact with the

termini of the nanotube electrodes and are embedded within the device.

Copper nanowires connect the device to an LCR meter in the laboratory.

LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.184

NATURE NANOTECHNOLOGY | VOL 6 | DECEMBER 2011 | www.nature.com/naturenanotechnology790

© 2011 Macmillan Publishers Limited. All rights reserved.

non-stretchable polyester substrate was approximately two timesgreater than that of the present device, as determined by the rela-tive increase in capacitance measured in pixels 4 mm from the siteof the load33.

Our device is less sensitive than another ‘skin-like’ devicereported in the literature33, but there are currently no other

devices that are both transparent and stretchable, and few havedemonstrated the ability to detect both pressure and strain4. Onedevice has been demonstrated based on nanowire field-effecttransistors that could detect a few kilopascals34, whereas anotherdevice with the geometry of a fishnet could undergo tensile defor-mations while detecting pressures on the order of 10 kPa (ref. 35).The pressures detectable by our devices, �50 kPa, correspondroughly to that of a firm pinch by opposing fingers. The architectureof the device, however, was not optimized.

Our devices were monolithically integrated, extremely mechani-cally compliant, physically robust and easily fabricated. The stretch-able, transparent nanotube electrodes were prepared withoutdispersion in an elastic matrix, without pre-straining the substrate,and patterned simply using stencil masks. In the future, it shouldbe possible to use these materials and principles to designorganic, skin-like devices with other human—and superhuman—characteristics36, such as the abilities to sense moisture, temperature,light6 and chemical and biological species37.

MethodsPreparation of substrates. PDMS (Dow Corning Sylgard 184; ratio of base tocrosslinker, 10:1 by mass) was mixed, degassed and poured against the polishedsurface of a silicon wafer bearing either a 300 nm thermal oxide or native oxidelayer. Before first use, the surfaces of the wafers were activated with oxygenplasma (150 W, 60 s, 400 mtorr) and passivated with the vapours of(tridecafluoro-1,1,2,2-tetrahydrooctyl)-1-trichlorosilane in a vacuum desiccator for≥4 h. Curing in an oven at 60 8C for 2 h produced PDMS membranes that were�0.3 mm thick. These membranes were cut into squares or rectangles with a razorblade with lengths and widths of 2–8 cm.

Preparation of carbon nanotube solution. Arc-discharge single-walled nanotubes(Hanwaha Nanotech Corp.) were ultrasonicated in N-methylpyrrolidone at 30%power for 30 min. The solution was then centrifuged for 30 min at 8,000 r.p.m. toremove large bundles, amorphous carbon or other contaminants. The top 75% ofthe solution was decanted for spray coating. The final solution had a concentrationof �150 mg ml21.

Spray coatings and dependence of contact angle on strain. Nanotubes were spray-coated using a commercial airbrush (Master Airbrush, Model SB844-SET). ThePDMS substrates were first activated with ultraviolet/ozone for 20 min, then held at180 8C on a hotplate, and the nanotubes were sprayed at a distance of �10 cm usingan airbrush pressure of 35 psi. We used a laser-cut PDMS membrane that had long,parallel rectangular apertures (to produce parallel lines) as a stencil mask. Multiplepasses of the airbrush (.100) were performed until the desired transparency wasreached. The patterned substrates were placed in a vacuum oven at 100 8C for 1 h toremove residual solvent.

We found that the surface of the ultraviolet/ozone-treated PDMS substratesbecame more hydrophobic with strain: when stretching from 0 to 60%, the watercontact angle increased from 70 to 908. Activation of the surface was necessary toform films in which the bundles of nanotubes were dispersed well.

Doping nanotube networks. F4TCNQ (TCI America) was dissolved to aconcentration of 0.4 mmol in chloroform by bath sonication for 45 min to 1 h. Theresulting bright yellow solution was filtered using a syringe filter before use. Afterfabrication, carbon nanotube networks were covered with sufficient solution to coverthe sample surface. The solution was left to sit for 60 s, and excess was then removedby spinning at 3,000 r.p.m. for 40 s. Samples were left to air dry for at least 30 minbefore measuring.

Fabrication of capacitive arrays. PDMS substrates patterned with nanotubes werestretched to 25% before laminating with one another. We mixed and degassedEcoflex 0010 silicone elastomer (Smooth-On 0010, TFB Plastics, 1:1 base tocrosslinker by volume) and spread it (using a piece of PDMS membrane as a‘paintbrush’) over the surface of one of the patterned substrates. We oriented asecond substrate, face down, perpendicular to the first, and pressed down. Weexpelled air bubbles and excess Ecoflex by rolling using a roll of tape. We placeddrops of EGaIn (Aldrich) on one of the two exposed termini of each line, placed acopper wire in each drop of EGaIn, and embedded the EGaIn drops with additionalEcoflex. Curing at 100 8C for 1 h produced monolithic arrays of capacitivepressure sensors.

Electrochemical, optical and sheet-resistance measurements. We measuredresistance versus strain of single- and multipixel devices by clamping the device intoa purpose-built, programmable stage to apply tensile strain. We measured resistanceand capacitance using an LCR (inductance, capacitance, resistance) meter (AgilentE498A precision LCR meter) interfaced with a custom LabView script. We measuredcapacitance versus strain by applying compressive force perpendicular to the device,

aHigh-contrast

Backlitb

c

d

ΔC/C

0

× 10−2

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0.0

Figure 5 | Images showing the characteristics of a 64-pixel array of

compressible pressure sensors. a, Photograph of the device, with enhanced

contrast to show the lines of nanotubes (scale bar, 1 cm). b, Photograph of

the same device reversibly adhered to a backlit liquid-crystal display. c, Map

of the estimated pressure profile over a two-dimensional area based on the

change in capacitance registered by a central pixel and its four nearest

neighbours when a pressure of 1 MPa is applied to the central pixel (scale

bar, 2 mm). d, Image of the device being deformed by hand.

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.184 LETTERS

NATURE NANOTECHNOLOGY | VOL 6 | DECEMBER 2011 | www.nature.com/naturenanotechnology 791

© 2011 Macmillan Publishers Limited. All rights reserved.

with the device placed between a programmable vertically movable stage and a forcegauge (Mark-10 model BG05) with a probe (area of contact defined by a square cutfrom a glass slide). In all cases, we used EGaIn to form deformable electrical contactsto the stretchable nanotube films.

We measured the optical transmission of the nanotube films using a Cary 6000ispectrophotometer. The reported values of transmission were taken at 550 nm.

We obtained measurements of sheet resistance using four collinear, equallyspaced probes connected to a Keithley 2400 sourcemeter.

Received 7 September 2011; accepted 27 September 2011;published online 23 October 2011; corrected online 28 October 2011

References1. LeMieux, M. C. & Bao, Z. N. Flexible electronics: stretching our imagination.

Nature Nanotech. 3, 585–586 (2008).2. Kim, B. Y. S., Rutka, J. T. & Chan, W. C. W. Current concepts: nanomedicine.

New Engl. J. Med. 363, 2434–2443 (2010).3. Ilievski, F., Mazzeo, A. D., Shepherd, R. F., Chen, X. & Whitesides, G. M. Soft

sobotics for chemists. Angew. Chem. Int. Ed. 50, 1890–1895 (2011).4. Cotton, D. P. J., Graz, I. M. & Lacour, S. P. A multifunctional capacitive sensor

for stretchable electronic skins. IEEE Sens. J. 9, 2008–2009 (2009).5. Sekitani, T. et al. Stretchable active-matrix organic light-emitting diode display

using printable elastic conductors. Nature Mater. 8, 494–499 (2009).6. Lipomi, D. J., Tee, B. C.-K., Vosgueritchian, M. & Bao, Z. N. Stretchable organic

solar cells. Adv. Mater. 23, 1771–1775 (2011).7. Ko, H. C. et al. A hemispherical electronic eye camera based on compressible

silicon optoelectronics. Nature 454, 748–753 (2008).8. Kim, D. H. et al. Dissolvable films of silk fibroin for ultrathin conformal bio-

integrated electronics. Nature Mater. 9, 511–517 (2010).9. Kim, R. H. et al. Waterproof AlInGaP optoelectronics on stretchable substrates

with applications in biomedicine and robotics. Nature Mater. 9, 929–937 (2010).10. Kim, D. H. et al. Epidermal electronics. Science 333, 838–843 (2011).11. Viventi, J. et al. A conformal, bio-interfaced class of silicon electronics for

mapping cardiac electrophysiology. Sci. Transl. Med. 2, 24ra22 (2010).12. Graz, I. M., Cotton, D. P. J. & Lacour, S. P. Extended cyclic uniaxial loading

of stretchable gold thin-films on elastomeric substrates. Appl. Phys. Lett. 98,071902 (2009).

13. Jones, J., Lacour, S. P., Wagner, S. & Suo, Z. G. Stretchable wavy metalinterconnects. J. Vac. Sci. Technol. A 22, 1723–1725 (2004).

14. Tahk, D., Lee, H. H. & Khang, D. Y. Elastic moduli of organic electronicmaterials by the buckling method. Macromolecules 42, 7079–7083 (2009).

15. Zhang, Y. Y. et al. Polymer-embedded carbon nanotube ribbons for stretchableconductors. Adv. Mater. 22, 3027–3031 (2010).

16. Kim, K. S. et al. Large-scale pattern growth of graphene films for stretchabletransparent electrodes. Nature 457, 706–710 (2009).

17. Avouris, P. Carbon nanotube electronics and photonics. Phys. Today 62,34–40 (2009).

18. Bae, S. et al. Roll-to-roll production of 30-inch graphene films for transparentelectrodes. Nature Nanotech. 5, 574–578 (2010).

19. Feng, C. et al. Flexible, stretchable, transparent conducting films made fromsuperaligned carbon nanotubes. Adv. Funct. Mater. 20, 885–891 (2010).

20. Hu, L. B., Yuan, W., Brochu, P., Gruner, G. & Pei, Q. B. Highly stretchable,conductive, and transparent nanotube thin films. Appl. Phys. Lett. 94,161108 (2009).

21. Yu, Z. B., Niu, X. F., Liu, Z. & Pei, Q. B. Intrinsically stretchable polymerlight-emitting devices using carbon nanotube-polymer composite electrodes.Adv. Mater. 23, 3989–3994 (2011).

22. Chun, K. Y. et al. Highly conductive, printable and stretchable composite films ofcarbon nanotubes and silver. Nature Nanotech. 5, 853–857 (2010).

23. Yu, C. J., Masarapu, C., Rong, J. P., Wei, B. Q. & Jiang, H. Q. Stretchablesupercapacitors based on buckled single-walled carbon nanotube macrofilms.Adv. Mater. 21, 4793–4797 (2009).

24. Bekyarova, E. et al. Electronic properties of single-walled carbon nanotubenetworks. J. Am. Chem. Soc 127, 5990–5995 (2005).

25. Hu, L. B., Hecht, D. S. & Gruner, G. Carbon nanotube thin films: fabrication,properties, and applications. Chem. Rev. 110, 5790–5844 (2010).

26. Nosho, Y., Ohno, Y., Kishimoto, S. & Mizutani, T. The effects of chemical dopingwith F(4)TCNQ in carbon nanotube field-effect transistors studied by thetransmission-line-model technique. Nanotechnology 18, 415202 (2007).

27. Khang, D. Y. et al. Molecular scale buckling mechanics in individual alignedsingle-wall carbon nanotubes on elastomeric substrates. Nano Lett. 8,124–130 (2008).

28. Kubo, M. et al. Stretchable microfluidic radiofrequency antennas. Adv. Mater.22, 2749–2752 (2010).

29. Cao, Q. & Rogers, J. A. Ultrathin films of single-walled carbon nanotubesfor electronics and sensors: a review of fundamental and applied aspects.Adv. Mater. 21, 29–53 (2009).

30. Jackman, R. J., Duffy, D. C., Cherniavskaya, O. & Whitesides, G. M. Usingelastomeric membranes as dry resists and for dry lift-off. Langmuir 15,2973–2984 (1999).

31. So, J. H. et al. Reversibly deformable and mechanically tunable fluidic antennas.Adv. Funct. Mater. 19, 3632–3637 (2009).

32. Dickey, M. D. et al. Eutectic gallium-indium (EGaIn): a liquid metal alloy for theformation of stable structures in microchannels at room temperature. Adv.Funct. Mater. 18, 1097–1104 (2008).

33. Mannsfeld, S. C. B. et al. Highly sensitive flexible pressure sensors withmicrostructured rubber dielectric layers. Nature Mater. 9, 859–864 (2010).

34. Takei, K. et al. Nanowire active-matrix circuitry for low-voltage macroscaleartificial skin. Nature Mater. 9, 821–826 (2010).

35. Someya, T. et al. Conformable, flexible, large-area networks of pressure andthermal sensors with organic transistor active matrixes. Proc. Natl Acad. Sci. USA102, 12321–12325 (2005).

36. Sokolov, A. N., Tee, B. C.-K., Bettinger, C. J., Tok, J. B.-H. & Bao, Z. N. Chemicaland engineering approaches to enable organic field-effect transistors forelectronic skin applications. Acc. Chem. Res. (in the press).

37. Roberts, M. E., Sokolov, A. N. & Bao, Z. N. Material and deviceconsiderations for organic thin-film transistor sensors. J. Mater. Chem. 19,3351–3363 (2009).

AcknowledgementsThis work was supported by a US Intelligence Community Postdoctoral Fellowship (toD.J.L.) and the Stanford Global Climate and Energy Program. B.C-K.T. was supported bythe Singapore National Science Scholarship from the Agency for Science Technology andResearch (A*STAR). The authors thank V. Ballarotto for helpful discussions and J.A.Bolander for writing code for the apparatus used for electromechanical measurements.

Author contributionsD.J.L. and Z.B. conceived the project. D.J.L., M.V. and B.C-K.T. performed and designedthe experiments. S.L.H. prepared the materials and developed the conditions used to dopethe nanotube films. J.A.L. deposited additional nanotube films. J.A.L. and C.H.F. performedexperiments on resistance versus strain. D.J.L., B.C-K.T., M.V., S.L.H. and Z.B. analysed thedata. D.J.L. wrote the paper. All authors discussed the results and commented onthe manuscript.

Additional informationThe authors declare no competing financial interests. Supplementary informationaccompanies this paper at www.nature.com/naturenanotechnology. Reprints andpermission information is available online at http://www.nature.com/reprints. Correspondenceand requests for materials should be addressed to Z.B.

LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2011.184

NATURE NANOTECHNOLOGY | VOL 6 | DECEMBER 2011 | www.nature.com/naturenanotechnology792

© 2011 Macmillan Publishers Limited. All rights reserved.