A stretchable carbon nanotube strain sensor for human ... - Nature

ARTICLES PUBLISHED ONLINE: 27 MARCH 2011 | DOI: 10.1038/NNANO.2011.36

A stretchable carbon nanotube strain sensor for human-motion detection Takeo Yamada1, Yuhei Hayamizu1, Yuki Yamamoto1, Yoshiki Yomogida1, Ali Izadi-Najafabadi1, Don N. Futaba1 and Kenji Hata1,2 * Devices made from stretchable electronic materials could be incorporated into clothing or attached directly to the body. Such materials have typically been prepared by engineering conventional rigid materials such as silicon, rather than by developing new materials. Here, we report a class of wearable and stretchable devices fabricated from thin films of aligned single-walled carbon nanotubes. When stretched, the nanotube films fracture into gaps and islands, and bundles bridging the gaps. This mechanism allows the films to act as strain sensors capable of measuring strains up to 280% (50 times more than conventional metal strain gauges), with high durability, fast response and low creep. We assembled the carbonnanotube sensors on stockings, bandages and gloves to fabricate devices that can detect different types of human motion, including movement, typing, breathing and speech.

C

onventional electronic devices, fabricated on rigid but brittle semiconductor wafers, have evolved through a drive towards miniaturization with a view to realizing faster, smaller and more integrated devices. An alternative approach to future electronics is to integrate the attributes of flexibility and stretchability to realize soft1–9 and human-friendly devices8–10. Stretchability—the ability to conform to and cover movable and arbitrarily shaped objects— could be exploited in the development of wearable devices8–10 that can be embedded into clothes and garments or even attached directly to the skin. Possible applications of this include the detection of human motion, monitoring personal health and therapeutics. Owing to the difficulties in developing stretchable electric materials, the current mainstream strategy in attempting to achieve stretchability is not to develop new materials, but instead is to engineer new structural constructs from established materials1. For example, ultrathin silicon structures formed into buckled geometries offer stretchability, and the strain applied to the device is absorbed by deformation of the silicon structures1. Consequently, the functional materials are exposed to a minimal amount of harmful strain. Various stretchable devices have been created using this approach, including conductors, transistors, diodes, photodetectors and integrated systems containing conventional rigid materials2,3, with the advanced functions and properties characteristic of modern semiconductor technology. A different approach is to assemble a device from stretchable materials. Examples of stretchable materials include polymer composites with conductive fillers8,11 and extremely thin metal films on stretchable polymer substrates12. In such stretchable devices, the functional materials themselves are directly exposed to strain and therefore stretched. This feature offers a unique opportunity to measure the strain-dependent change in device performance to monitor motion, for example of the human body. Here, we introduce a new type of stretchable electric nanomaterial consisting of aligned single-walled carbon nanotube (SWCNT) thin films13–15 that deform when stretched in a manner similar to the structural deformation of a string cheese when peeled. In this way we realized a novel strain sensor that can measure and withstand strain up to 280%, with high durability (10,000 cycles at

150% strain), fast response (delay time, 14 ms) and low creep (3.0% at 100% strain). These important features allow the material to be used to precisely monitor large-scale and rapid human motion, as was demonstrated by embedding various strain sensors into clothing worn over the skin then using it to detect movement, typing, breathing and phonation (speech).

Carbon nanotube film strain sensor Figure 1a schematically illustrates the key processes in fabricating and operating a SWCNT film strain sensor. Vertically aligned and very sparse (3–4% occupancy) SWCNT thin films (height, 1 mm; thickness, 6 mm; length, 16 mm) were first grown from patterned catalysts using water-assisted chemical vapour deposition13. To make long films of arbitrary length, films were individually removed and laid side by side, with a 1 mm overlap, onto a flat elastomeric dog-bone-shaped substrate (poly(dimethylsiloxane), PDMS; thickness, 1 mm), with the alignment of the SWCNTs arranged perpendicular to the strain axis (Fig. 1b). For each iteration, the film was wet with a droplet of isopropyl alcohol, which flattened the film (thickness, 400 nm) to the substrate in a manner similar to deflating an air mattress; this allowed the SWCNTs to be packed into a highly densely packed solid form (density, 0.46 g cm23; occupancy, 42%; SWCNT spacing, 4.1 nm)14,15. This process resulted in the development of a strong van der Waals contact with the substrate, achieved without any additional mechanical pressure. The adhesion strength was measured as 12 N cm22 and was sufficient to bear large strain. Representative resistivity–strain data recorded for the device in Fig. 1b showed a monotonic increase up to 280% strain (strain speed, 1 mm min21) (Fig. 1c), at which point the PDMS substrate ruptured. This monotonic increase in resistivity with strain demonstrates the potential use of this device as a gauge to measure strains much higher than the 5% limit of conventional metal strain gauges (Fig. 1c). Strain gauges register resistance changes when stretched, and are used to measure acceleration, pressure, tension and strain. The most commonly used strain gauge consists of a non-stretchable metal foil, which, on stretching, deforms to become narrower and longer, resulting in an increase in resistance16. The use of a

1

Nanotube Research Center, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, 305-8565, Japan, 2 Japan Science and Technology Agency (JST), Kawaguchi, 332-0012, Japan. * e-mail: [email protected]

296

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

NATURE NANOTECHNOLOGY

ARTICLES

DOI: 10.1038/NNANO.2011.36

a

Strain sensor

SWCNT films SWCNT films and in IPA stretchable substrate

Sensor under strain

b

Sensor under strain

c

500

10

40 35

100 40 38.4 36.8 35.2 33.6 0.5 1.5 2.5 3.5 4.5 5.5

30 25 20 15

4

Aligned SWCNTs Random SWCNTs Metal strain gauge

100 0 0

300

3 8 2 1

2

40

65 30 60 20 0

50

11 5 10

0 10 20

4 9 3 8 2 7 4.6

200

40

i

20 15

4.7

4.8 4.9 Time (s)

5.0

90

Sensor 3

25 Sensor 2

20 15 10

Sensor 1

5

10th 100th 1,000th 10,000th

0

85

PDMS packaging

10

SWCNTs 80

5

Resistance (k )

Relative change in resistance (%)

35 30

5

6

h

55

100 150 Strain (%)

4

Strain (%)

70

30

3 Time (s)


75

0

6

4

80

10

5

9

0

60

20

4

5

Resistance (k )


3 Time (s)

7

50

e

2

10


d

100 200 Strain (%)

1

11

8

200

20

6

g

0 0


300

40

88

0 0

5

60

Strain (%)


400

80

112 104 96

Strain (%)

Aligned direction

f Relative change in resistance (%)

SWCNT films

PDMS Electrode

0 50

100 150 Strain (%)

200

0

40

80 120 Strain (%)

160

Figure 1 | SWCNT-film strain sensor. a, Key steps in fabricating the SWCNT strain sensor. b, Photograph of the SWCNT-film strain sensor under strain. c, Relative change in resistance versus strain for the strain sensor (aligned SWCNTs, red), randomly oriented SWCNTs (blue) and conventional metal thin film (black). Inset: close-up of the low-strain region. d, Relative change in resistance for the initial loading (red) and unloading (blue) cycle. Inset: enlarged unloading plot. e, Relative change in resistance versus strain for multiple-cycle tests: 10 (red), 100 (blue), 1,000 (green) and 10,000 (black) cycles at 5–100% (sensor 1), 5–150% (sensor 2) and 5–200% (sensor 3) strain. The baselines of sensors 2 and 3 are raised by 5 and 10%, respectively. f, Relative change in resistance (blue) in response to a 5–100% step function of mechanical strain. Inset: close-up of the overshoot. g, Relative change in resistance (blue) during 2.5 Hz frequency cycling between 2.0 and 5.4% strain (red). h, Close-up of the final cycle of g. i, Initial loading (red) and unloading (blue) of relative change in resistance versus strain for the packaged sensor. Inset: image of the packaged sensor structure.

non-stretchable metal foil limits device performance, highlighting that the high deformability of SWCNT films results from its fibre structure, as discussed later. The resistivity behaviour observed during the first loading–unloading cycle and subsequent cycles

(strain speed, 6 mm s21) differed in the loading and unloading phases (Fig. 1d). The unloading behaviour (Fig. 1d, blue line) was characterized by two linear regions (strain of 0 to 40% and 60 to 200%) with different slopes. The slope reflects the gauge factor


297

ARTICLES a

NATURE NANOTECHNOLOGY 0%

b

c

25%

50%

d

75%

Island

e

DOI: 10.1038/NNANO.2011.36

100%

Gap

Initial strain

g

f

Buckling

is

in

a Str

Island

Suspended SWCNT bundle

Gap Island Buckling

i Distance ( m)

h

Gap

ax

Island width Gap width

8

j

6 4

k

2 0

Gap

0

20 40 60 80 100 Strain (%)

Figure 2 | Fracturing mechanism of film. a–e, Images of the SWCNT film on initial loading. Scale bar, 100 mm. f, SEM image of the fractural structure of the SWCNT film at 100% strain. Scale bar, 5 mm. Inset: three-dimensional image at 100% strain. g, Low-resolution SEM image of homogeneous fracturing of the SWCNT film. Scale bar, 50 mm. h, SEM image of the suspended bundles. Scale bar, 1 mm. i, Average island width (blue) and gap (red) versus strain for initial loading. Error bars indicate standard deviation. j,k, Paper model of the unstrained and strained states, respectively.

(the sensitivity of the SWCNT film to strain), which is defined as (dR/R)/(dL/L), where R is resistance and L is length. The gauge factors were calculated to be 0.82 (0 to 40% strain) and 0.06 (60 to 200%); in comparison, conventional metal gauges have a factor of 2.0 (5% maximum strain)16 and polymer composites (thermal plastic elastomer) with 50 wt% carbon black a factor of 20 (80% maximum strain)8. Our carbon-nanotube strain sensor exhibited excellent durability and stability, even at high strain levels (100, 150 and 200% at a strain speed of 6 mm s21). At 100 and 150% strain, the strain sensor electrical response remained nearly unchanged after 10,000 cycles (Fig. 1e). At 200% strain, the sensor was stable for 3,300 cycles, until the substrate ruptured, showing that the performance was limited by the substrate. Additional important advantages of the carbon-nanotube strain sensor include its low creep and fast response. The sensor was cycled between 5 and 100% strain at a speed of 10.6 mm s21 and with a recovery time of 5 s (Fig. 1f, red line). The response of the carbon-nanotube strain sensor to this strain (Fig. 1f, blue line) was fast, with a low overshoot of 3.0% and recovery time of 5 s. This is markedly different from the 8.8% overshoot and more than 100 s recovery time observed for polymer composites with conductive fillers, even with a three times lower strain speed8. To evaluate the delay time, the nanotube strain sensor was subjected to a sinusoidal strain. The electrical delay was evaluated to be 14 ms (Fig. 1g,h; Supplementary Fig. S1). To the best of our knowledge, this is the only reported delay time on the order of milliseconds in this kind of large strain gauge to date. Several tests were performed to clarify the selectivity of the strain sensor to other types of deformations, such as twist and 298

compression. The strain sensor (length, 5 mm) was first twisted while measuring resistivity (Supplementary Fig. S2). The relative change in resistance at 908 twist was 0.5%, corresponding to strain of 2%. The strain sensor was then subjected to normal mechanical strain across the sensor face using a 1-cm-radius pressure head and a static mechanical analyser to load compressive stress while measuring the resistivity (Supplementary Fig. S2). The relative change in resistance at 30% compression was 1%, corresponding to 3% strain. These results demonstrate that the strain sensor has a high selectivity against other types of deformation. Further tests were carried out to investigate environmental effects such as temperature and humidity. First, by its nature, the strain sensor is sensitive to temperature, because the expansion/shrinkage of the substrate caused by variation in temperature will be detected as a strain by the sensor. When the temperature was raised from room temperature to 50 8C, the relative change in resistance was 6%. This effect was observed as a slow drift in resistivity that can be corrected practically using a.c. measurements or, more precisely, compensated directly using a temperature monitor. Second, any environmental effect that influences the conductivity of carbon nanotubes would also influence the strain sensor. For example, exposure to gases is well known to dope carbon nanotubes and change their conductivity. When the strain sensor was exposed to an exhaled breath (Supplementary Fig. S3), the resistivity changed by 0.6%, correspondsing to 2% strain. Packaging is therefore important in reducing the sensitivity of the strain sensor to environmental effects and to prevent damage from abrasion. To address this, the strain sensor was sealed with a PDMS coating. Importantly, when sealed, the behaviour of the strain sensor


NATURE NANOTECHNOLOGY a

ARTICLES

DOI: 10.1038/NNANO.2011.36

100% b

50%

c

0%

d

50%

Gap

100%

Island

Strain unloading

Strain loading

g

f 6

R1

5

R2

4

RC

Distance ( m)

e

3 2 1

R2

Island width Gap width

R1

0 0

20

40 60 80 Strain (%)

100

Figure 3 | Properties of SWCNT film. a–e, Film surface morphology during the unloading–loading cycle. Scale bar, 100 mm. f, Average island width (blue) and gap (red) versus strain for cycling following the conditioning step. Error bars indicate standard deviation. g, Model of the basic circuit of the film.

(Fig. 1i) was similar to that of the unsealed sensor, with only a slight decrease in the gauge factor (0.05). Moreover, the sealed strain sensor showed improved linearity, with a linear increase in resistivity observed up to 150% strain. There have been several reports of the individual strengths and advantages of other stretchable materials, but no material has achieved this level of stretchability and durability, as well as a fast response and low creep. For example, a polymer composite (thermal plastic elastomer) with conductive material (carbon black, 50 wt%) has been assembled into textiles to measure large strains (80%)8. This composite material has a large interfacial area and therefore friction, leading to long delays (recovery time above 100 s, versus 5 s for the present arrangement) and creep (8.8% versus 3.0%). Thin metal films on flexible and stretchable polymer substrates rupture into connected islands, and have been used as strain sensors up to 60% strain12. However, their delay time was over 40 s at 0.5% strain12. Several reports have used composites of carbon nanotubes and polymers as strain sensors, and the results are summarized in Supplementary Table 1. Although some of those sensors had a large gauge factor (15), they only tolerated a small maximum strain (1%). A multiwalled carbon-nanotube forest was infiltrated with polyurethane (PU) to make a conducting polymer composite that was more stretchable (1,000%) than PU alone17. However, the large drift in resistance over repeated loading–unloading cycles and probable large delay time owing to the large interfacial area of the carbon nanotubes and polymer would hinder its use as a strain sensor17. No reported sensors have shown a performance level comparable to the sensor reported here in terms of maximum strain, durability, creep and delay time. The nanotube strain sensor can be regarded as a cross between thin-film-based strain sensors and composite strain sensors, where the conducting filler material is not infiltrated, but is instead coated onto the surface of the polymer, thereby combining the strengths of both systems. This configuration provides a minimum interfacial area between the nanotubes and the polymer, thereby minimizing the friction18 that causes delay, creep and durability. In fact, our preliminary analysis showed that delay and creep were dominated by the substrate properties (Supplementary Fig. S4), suggesting that the nanotube strain sensor could be further improved. In contrast to traditional stretchable conducting materials, these results demonstrate that the nanotube strain sensor has the potential to precisely monitor rapid and large-scale human motion.

Fracturing mechanism of carbon nanotube film To understand the underlying mechanism of the film stretchability, we examined the structural change in the SWCNT film under different levels of strain (Fig. 2a–e; Supplementary Movie 1). On initial stretching, irreversible fracturing throughout the film created gaps and islands; with further strain, their numbers and also the gap widths increased (Fig. 2f–i), explaining the observed linear increase in resistivity (Fig. 1d). Uniformity of the SWCNT film was vital for homogeneous fracturing throughout the film (Fig. 2g). In addition, the film showed buckling parallel to the strain axis, because it followed the same positive Poisson’s ratio as the substrate; that is, the substrate width narrowed as strain was applied (Fig. 2f ). Importantly, suspended SWCNT bundles bridging the gaps (Fig. 2h; with an appearance similar to peeled string cheese) were created, preventing film rupturing. This highlights the importance of the alignment of the SWCNTs in allowing exceptional strain tolerance compared with randomly oriented vacuum filtrated SWCNT buckypaper (5% failure strain). Lateral interconnections between the SWCNTs were essential to create the suspended bundles, and millimetre-scale SWCNTs produced durable and elongated bundles spanning wide gaps. The islands, in contrast, acted as anchors to prevent film detachment. As demonstrated using a stretchable paper model (Fig. 2j,k), the mechanism of stretchability is analogous to the structural deformation of the open-mesh geometries used to wrap three-dimensional objects, such as eggs4. When the paper film was stretched, open holes deformed to allow stretching (Fig. 2k), while the strips acted as bending units. This process was downsized by a factor of 2,000 for the SWCNT film, with suspended bundles acting as the bending units.

Repeated loading–unloading cycle properties The structural deformation (Fig. 3a–c) observed during the first unloading phase and subsequent loading–unloading cycles differed significantly. During the subsequent phases (Fig. 3c–e), new gaps and islands were neither created nor annihilated, as demonstrated by the unvarying widths of the islands with strain (Fig. 3f ). Instead, all of the strain was absorbed by a reversible opening and closing of the gaps (Fig. 3a–e; Supplementary Fig. S5, Movie 1). The reversible fracturing of the SWCNT film not only explained the extreme level of stretchability, but was also key to the exceptional durability, as demonstrated by the 10,000 cycle durability test (Fig. 1e), which showed little degradation. The resistance of the film increased monotonically (Fig. 1e) but not linearly with strain,


299

ARTICLES

NATURE NANOTECHNOLOGY Relative change in resistance (%)

a

b Strain sensor Bandage

d

2

SWCNT rubber PDMS glue SWCNTs Support electrode

0

DOI: 10.1038/NNANO.2011.36

−2

PDMS

Strain sensor

0

20 40 Time (s)

60


c

0

Stocking

−2

f Relative change in resistance (%)

5%

e

120

Extending

2s

Flexing

Marching Squatting Jumping Time (s)

Jumping from squatting

Strain sensor

Glove

g


40 80 Time (s)

Little Ring Middle Index Thumb 100%

0

0

Time (s)

20

Figure 4 | Stretchable wearable devices. a,d,f, Photographs of a bandage strain sensor (a), a strain sensor fixed to a stocking (d) and a data glove (f). Inset to a: Photograph of the sensor adhered to the throat. Inset to d: close-up of the device. b,c,e,g, Relative changes in resistance versus time for breathing, phonation (speech), knee motion and data glove configurations, respectively.

and was characterized by two consecutive regions with different gauge factors. To understand this resistivity behaviour, we developed a simple model to describe the resistance of the fractured film (Fig. 3g): R=

R1 Rc + 2R1 R2 + R2 Rc R1 + 2Rc + R2

(1)

where R1 , R2 and Rc are the resistances of the island, gap and suspended bundle, respectively. Physically, suspended bundles are analogous to wires, and Rc (bridge elongation) should increase linearly with strain. Therefore, at large strain, when R2 ≫ R1 and Rc , equation (1) becomes R ¼ 2R1 þ Rc . R1 , and Rc can be determined by fitting to experimental data (Supplementary Fig. S6) (R1 ¼ 37.6 kV; Rc ¼ 28.0 × strain (%) V). Thus, by knowing R1 and Rc , the value of R2 is shown to increase exponentially with strain (Supplementary Fig. S6) (R2 ¼ 147.9 × exp(0.0515 × strain (%)) kV). The good agreement between model and experiment (Supplementary Fig. S6) allows us to gauge the resistance to absolute strain. From these results, we interpret the double gauge factor behaviour of the resistance of the strain sensor as deriving from a shift in the dominant contributions from gap opening (R2) under low strain, to bridge elongation (Rc) under high strain.

Application in human motion detection To demonstrate the potential of the SWCNT films in wearable devices8–10, we fabricated a stretchable human motion detector by connecting stretchable electrodes to the films and assembling them on bandages and clothing. To avoid mechanical failure at the junction between the stretchable SWCNT film and rigid electrodes, we used SWCNT conductive rubber paste5,6 to fabricate the stretchable electrodes and a PDMS rubber glue (SH-780) to assemble the components onto the target material (bandage, clothing, and so on). The SWCNT conductive paste was fabricated by dispersing highly elastic fluorinated copolymer rubber into a SWCNT gel. This gel was composed of millimetre-long SWCNTs and an ionic liquid mixed by a high-pressure jet-milling homogenizer. This SWCNT rubber paste was painted on the ends of the SWCNT film and dried to form an elastic conductor. Finally, the PDMS glue was used to cover the electrodes, both for reinforcement and to fix the device onto a commercial adhesive bandage. 300

Adhesive bandages are used daily to protect small wounds and adhere securely to the body and conform to every contortion of the skin. As a result of the stretchable device architecture, the skin, bandage and SWCNT film device behave as a single cohesive stretchable object, so deformation of the skin can be monitored directly and precisely using the SWCNT film (Fig. 4a). When fixed to the chest, respiration could be monitored by the upward and downward slopes of the relative resistance associated with inhalation and exhalation (chest expansion and contraction) (Fig. 4b). In contrast, when attached to the throat (Fig. 4a, inset), the device monitored phonation (speech) by detecting motion of the laryngeal prominence (the Adam’s apple) (Fig. 4c). Such devices might be useful in a breathing monitor for the early detection of sudden infant death syndrome (SIDS) in sleeping infants, alerting parents to any potential problems19. In the basic act of walking, the skin on the feet, waist and joints repeatedly stretches and contracts by as much as 55% (Supplementary Fig. S7), which already exceeds the limits (5%) for conventional strain sensors. To detect large-scale human motion, we seamlessly connected small films to fabricate a large SWCNT strain sensor with extended sensing area. Eight films were connected together to form a 10 cm × 1 mm SWCNT film assembled on a commercial stocking (Fig. 4d) over the knee joint. The large SWCNT film was necessary to detect and distinguish every movement of the knee. As the knee joint moves in one direction (as well as swivelling on its axis), the knee constantly rolls and glides during movement, so the deformation site of the skin is constantly varying. Although it was made from just one sensor, the device could easily detect, and also discriminate, various human motions related to the extension and flexion of the knee, including bending, marching, squatting and jumping, and combinations of these (Fig. 4e; Supplementary Movie 2). For example, the complex motion of explosive jumping using a squatting motion could be clearly determined from the initial knee flexing (upward slope), quick knee extension (downward slope), slight flexure from the landing (second upward slope) and recovery (final downward slope). One advantage of using clothing-integrated devices is the option for repeatable and sharable use of the sensor. In addition, this device does not restrict motion, in contrast to conventional rotary encoders, and would therefore be beneficial for humanfriendly rehabilitation20.


NATURE NANOTECHNOLOGY

DOI: 10.1038/NNANO.2011.36

Integration of the SWCNT strain devices creates a system with which to examine the configuration of the human body, as demonstrated by a data glove made from five independent SWCNT strain sensors assembled on a single glove (Fig. 4f ). A data glove is an interactive device, resembling a glove normally worn on the hand, which facilitates fine-motion control in robotics and virtual reality. Our data glove could detect the motion of each finger individually and precisely (Fig. 4g; Supplementary Movie 3), and the output of each gauge could be measured to assess the hand configuration. Conventional data glove systems use optical fibres or metal strain gauges (cyber glove) as sensing elements21. Our data glove is lighter, simpler, allows integration of more sensors than the complex optical fibre system, and does not limit any range of motion of the hand, as does the metal-strain-gauge system. This device might be used as a master-hand to control a remote slave robot to remotely perform surgical procedures (telesurgery)22 or to increase demining safety and speed23.

Conclusions Although the functions of the devices presented here were simple compared to the wearable systems made from commercial devices24, we have presented a route to materials and structures, developed through nanotechnology, that can be used to develop human-friendly devices with realistic functions and abilities that would not be feasible by mere extension of conventional technology. Our research suggests devices that can act as part of human skin or clothing, and can therefore be used ubiquitously. We believe that such devices could eventually find a wide range of applications in recreation, virtual reality, robotics and health care.

Methods SWCNT film devices for strain sensor characterization were fabricated on a dogbone-shaped backing structure made of PDMS (SH9555, Dow Corning Toray). The backing structure was 1 mm thick, with a width of 25 mm and an overall length of 65 mm. Strain property characterizations of the devices were performed on a computercontrolled, home-made actuating unit located on an optical bench. Surface structure evaluations of the SWCNT film (islands, gaps and suspended bundle structure) were performed using a laser microscope (VK-9700, Keyence) and a scanning electron microscope (S-4800, Hitachi). Fabrication of the wearable devices followed the same procedure as that for the SWCNT devices, using 1-mm-thick rectangular PDMS (SILPOT 184, Dow Corning Toray) backings of appropriate size. Before setting the SWCNT films, support electrodes of Ti(3 nm)/Au(100 nm)/Ti(10 nm) were deposited at both ends of the substrate. After setting and drying the films, lead wires were connected to the support electrodes using carbon nanotube rubber6 in place of the rigid silver paste, and the rubber was coated with PDMS glue (SH-780, Dow Corning Toray). Finally, the device was glued to the target clothing. For human-motion sensing, the lead wires of the sensor(s) were connected to a potentiostat (VMP3, Princeton Applied Research).

Received 3 August 2010; accepted 18 February 2011; published online 27 March 2011

References 1. Khang, D. Y., Jiang, H., Huang, Y. & Rogers, J. A. A stretchable form of singlecrystal silicon for high-performance electronics on rubber substrates. Science 311, 208–212 (2006). 2. Kim, D. H. et al. Stretchable and foldable silicon integrated circuits. Science 320, 507–511 (2008).

ARTICLES

3. Kim, D. H. & Rogers, J. A. Stretchable electronics: materials strategies and devices. Adv. Mater. 20, 4887–4892 (2008). 4. Someya, T. et al. Conformable, flexible, large-area networks of pressure and thermal sensors with organic transistor active matrixes. Proc. Natl Acad. Sci. USA 102, 12321–12325 (2005). 5. Sekitani, T. et al. A rubberlike stretchable active matrix using elastic conductors. Science 321, 1468–1472 (2008). 6. Sekitani, T. et al. Stretchable active-matrix organic light-emitting diode display using printable elastic conductors. Nature Mater. 8, 494–499 (2009). 7. Urdaneta, M. G., Delille, R. & Smela, E. Stretchable electrodes with high conductivity and photo-patternability. Adv. Mater. 19, 2629–2633 (2007). 8. Mattmann, C., Clemens, F. & Tro¨ster, G. Sensor for measuring strain in textile. Sensors 8, 3719–3732 (2008). 9. Munro, B. J., Campbell, T. E., Wallace, G. G. & Steele, J. R. The intelligent knee sleeve: a wearable biofeedback device. Sens. Actuat. B 131, 541–547 (2008). 10. Scilingo, E. P., Lorussi, F., Mazzoldi, A. & Rossi, D. D. Strain-sensing fabrics for wearable kinaesthetic-like systems. IEEE Sens. J. 3, 460–467 (2003). 11. Cochrane, C., Koncar, V., Lewandowski, M. & Dufour, C. Design and development of a flexible strain sensor for textile structures based on a conductive polymer composite. Sensors 7, 473–492 (2007). 12. Lacour, S. P., Jones, J., Wagner, S., Li, T. & Suo, Z. G. Stretchable interconnects for elastic electronic surfaces. Proc. IEEE 93, 1459–1467 (2005). 13. Hata, K. et al. Water-assisted highly efficient synthesis of impurity-free singlewalled carbon nanotubes. Science 306, 1362–1364 (2004). 14. Futaba, D. N. et al. Shape-engineerable and highly densely packed single-walled carbon nanotubes and their application as super-capacitor electrodes. Nature Mater. 5, 987–994 (2006). 15. Hayamizu, Y. et al. Integrated three-dimensional microelectromechanical devices from processable carbon nanotube wafers. Nature Nanotech. 3, 289–294 (2008). 16. Dobie, W. B. & Isaac Peter, C. G. Electric Resistance Strain Gauges (English Universities Press Limited, 1948). 17. Shin, M. K. et al. Elastomeric conductive composites based on carbon nanotube forests. Adv. Mater. 22, 2663–2667 (2010). 18. Qiang, L. et al. Improved electrical resistance–pressure strain sensitivity of carbon nanotube network/polydimethylsiloxane composite using filtration and transfer process. Chin. Sci. Bull. 55, 326–330 (2010). 19. Fleming, P. J. et al. Interaction between bedding and sleeping position in the sudden infant death syndrome: a population based case-control study. Br. Med. J. 301, 85–89 (1990). 20. Nikitczuk, J., Weinberg, B. & Mavroidis, C. Control of electro-rheological fluid based resistive torque elements for use in active rehabilitation devices. Smart Mater. Struct. 16, 418–428 (2007). 21. Dipietro, L., Sabatini, A. M. & Dario, P. A survey of glove-based systems and their applications. IEEE Trans. Syst. Man. Cybern. 38, 461–482 (2008). 22. Hanly, E. J. & Talamini, M. A. Robotic abdominal surgery. Am. J. Surg. 188, 19S–26S (2004). 23. Wojtara, T. et al. Hydraulic master–slave land mine clearance robot hand controlled by pulse modulation. Mechatronics 15, 589–609 (2005). 24. Picard, R. W. & Healey, J. Affective wearables. Pers. Ubiquitous Comput. 1, 231–240 (1997).

Acknowledgements The authors thank Y. Yamada and T. Toida for their assistance. The authors also acknowledge partial support from Core Research for Evolutional Science and Technology (CREST) of the Japan Science and Technology Agency (JST).

Author contributions T.Y., Y.H. and K.H. conceived and designed the experiments. T.Y. and Yu.Y. performed the experiments. D.F. contributed to materials preparation, Yo.Y. and A.I. contributed to device demonstration. T.Y. and K.H. co-wrote the paper.

Additional information The authors declare no competing financial interests. Supplementary information accompanies this paper at www.nature.com/naturenanotechnology. Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/. Correspondence and requests for materials should be addressed to K.H.


301