Within Touch of Artificial Skin

As humans, we interact with our immediate environment through our senses — sight, sound, smell, taste and touch. Emulation of the senses by electronic means has long been a grand challenge of artificial intelligence and is pivotal in the development of accessible and natural interfaces between man and machine. Sight and sound are the simplest technologically, and the past decades have seen significant developments in image acquisition and processing technologies, in addition to voice synthesis and recognition. From an electronic standpoint, smell and taste are one and the same, but despite significant advances in electronic nose1 and chemical sensor2 technologies, the sensitivity and discrimination levels of these engineered systems fall short of the performance of their human counterparts. Touch also remains stubbornly difficult to mimic. The difficulty is not simply to identify transduction mechanisms that can detect mechanical resistance or static mass loads; touch emulation necessitates the development of high spatial-resolution, pressure-sensitive artificial skins capable of discriminating between local stimuli on a textured surface. For example, by applying a pressure of 10 kPa over a 1 cm2 contact area, the human touch can typically detect local roughness variations with a spatial resolution of 50 μm. Coupled with the fact that the softest touch corresponds to a mass-loading sensitivity of better than 0.1 g per mm2 (or about 1 kPa), this highlights the very real challenge of developing touch technology that can compete with the performance in humans. As now reported in Nature Materials, separate research groups have addressed this problem in two distinct ways3,4. Both employ an active matrix array of transducers using flexible materials. Flexibility is desirable because it enables the fabrication of transducer arrays that can conform to curved surfaces, which is essential if these engineered materials are to serve as artificial skins for prosthetic devices or in applications where a high degree of spatial resolution is required. Where the groups differ is in their approach to the transduction mechanism and the types of substrate used. Ali Javey and co-workers3 used Ge/Si-nanowire-array field-effect transistors (FETs) laminated on a flexible polyimide substrate with a pressuresensitive rubber layer that acts as a tunable resistor in series with the nanowire FET (Fig. 1). On the other hand, Zhenan Bao and collaborators4 microstructured polydimethylsiloxane (PDMS) films to produce pressure-sensitive capacitor arrays that are integrated into the gate dielectrics of an organic FET array (Fig. 2). Both report pressure sensors with response times of less than 100 ms and a dynamic range of 0.5–20 kPa or better, and each represents a significant advance in the state of the art. By tailoring the microstructured PDMS film, it is possible to achieve static load sensitivities as low as 3 Pa, and ultrafast millisecond response times4. In the approach of Bao and colleagues4, array integration involves laminating the PDMS film onto a non-flexible silicon substrate, so the overall conformability is lost. However, these authors also reported a proof-ofconcept flexible capacitive matrix-type pressure sensor (Fig. 2b). In contrast, the nanowire-on-polyimide approach ensures flexibility, and Javey and collaborators have FLEXIBLE ELECTRONICS