Silicon nanomembranes enable fingertip electronics
(Authors contributed equally to this work.)
1. Department of Materials Science and Engineering, Frederick Seitz Materials Research Laboratory, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA
2. Department of Mechanical Engineering and Department of Civil and Environmental Engineering, Northwestern University, Evanston, IL 60208, USA
3. Faculty of Infrastructure Engineering, Dalian University of Technology, Dalian 116024, People’s Republic of China
Abstract
We describe the use of semiconductor nanomaterials, advanced fabrication methods and unusual device designs for a class of electronics capable of integration onto the inner and outer surfaces of thin, elastomeric sheets in closed-tube geometries, specially formed for mounting on the fingertips. Multifunctional systems of this type allow electrotactile stimulation with electrode arrays multiplexed using silicon nanomembrane (Si NM) diodes, high-sensitivity strain monitoring with Si NM gauges, and tactile sensing with elastomeric capacitors.
Analytical calculations and finite element modeling of the mechanics quantitatively capture the key behaviors during fabrication/assembly, mounting and use. The results provide design guidelines that highlight the importance of the NM geometry in achieving the required mechanical properties. This type of technology could be used in applications ranging from human–machine interfaces to ‘instrumented’ surgical gloves and many others.
1. Introduction
Electrotactile stimulators and tactile sensors are of interest as bi-directional information links between a human operator and a virtual environment, in a way that could significantly expand function in touch-based interfaces to computer systems, with applications in simulated surgery, therapeutic devices, robotic manipulation, and others [1–5]. Electrotactile stimulation allows information to be presented through the skin, as an artificial sensation of touch, commonly perceived as a vibration or tingling feeling [6, 7]. Such responses are manifested through the excitation of cutaneous mechanoreceptors as a result of passage of a suitably modulated electrical current into the tissue [8]. Developed originally in the 1950s and further advanced in the 1970s, electrotactile stimulation has been traditionally explored for programmable braille readers and displays for the visually impaired as well as for balance control in individuals who suffer from vestibular disorders [5, 9–12]. Tactile sensors, on the other hand, measure the pressure created by physical contact, in a way that provides complementary information for potential use in feedback loops with the electrotactile process. Additional classes of sensors that can be important in this context include those for motion and temperature.
Incorporation of such technologies into a conformal, skin-like device capable of intimate, non-invasive mounting on the fingertips might, therefore, represent a useful achievement. Recent advances in flexible and stretchable electronics create opportunities to build this type of device [13–17]. Here we report materials, fabrication strategies and device designs for ultrathin, stretchable silicon-based electronics and sensors that can be mounted on the inner and outer surfaces of elastomeric closed-tube structures for integration directly on the fingertips. The active components and interconnects incorporate advanced mechanics designs, capable of accommodating large strains induced not only by natural deformations of the tubes during use, but also during a critical step in the fabrication process in which the tubes, specially formed to match the shapes of fingertips, are flipped inside out. This ‘flipping-over’ process allows devices initially mounted on the outer surface of the tube to be reversed to the inner surface, where they can press directly against the skin when mounted on the fingers. Analytical calculations and finite element modeling (FEM) provide quantitative insights into design layouts that avoid plastic deformation or fracture.
We demonstrate these concepts in multifunctional fingertip devices that include electrotactile electrode arrays multiplexed with Si nanomembrane (NM) diodes, strain sensors based on Si NM gauges, and tactile sensor arrays that use capacitors with low modulus, elastomeric dielectrics.
2. Experiments
Figure 1 schematically illustrates the steps for integrating devices based on Si NMs in stretchable, interconnected geometries with elastomeric substrates, following adapted versions of procedures described elsewhere [13, 18]. The fabrication uses a Si wafer with a 100 nm thick coating of polymethylmethacrylate (PMMA) as a temporary substrate for the initial parts of the process.
Figure 1: Schematic illustration of the process for transfer printing an interconnected device structure from a substrate on which it is fabricated to an elastomeric sheet. (a) Interconnected sensors and electronics formed on a silicon wafer in an open mesh geometry are lifted onto the surface of a PDMS slab (i.e. stamp); (b) The back side of the mesh and the supporting PDMS stamp are coated with a thin layer of SiO2 and then pressed onto an elastomeric sheet (Ecoflex);(c) Removal of the PDMS completes the transfer; (d) Materials legend.
Figure 2: The process for fabricating a multiplexed array of electrotactile stimulators in a stretchable, mesh geometry on the inner surface of an elastomeric finger-tube.
(a) Casting and curing an elastomer precursor on the finger of a model hand yields a thin (∼ 500µm thick), closed-form membrane, i.e. a finger-tube.
A layer of polyimide (PI, 1.25 µm thick) formed by spin coating a poly (amic acid) precursor and baking in an inert atmosphere at 250°C serves as the support for the devices. Electronically active materials are deposited (e.g. metallization) or transfer printed (e.g. Si NMs) onto the PI and patterned by photolithography and etching. Another layer of PI (1.25 µm thick) spin cast and cured on top of the device layers provides encapsulation and locates the devices near the neutral mechanical plane (NMP). Next, patterned reactive ion etching through the entire multilayer stack (i.e. PI/devices/PI) defines an open mesh structure. This same process removes the PI in the regions of the electrotactile stimulation electrodes, to allow direct contact with the skin. Immersion in an acetone bath washes away the underlying PMMA, thereby allowing the entire mesh to be lifted off, in a single piece, onto the surface of a flat slab of polydimethylsiloxane (PDMS), using procedures described previously [19, 20]. Evaporation of a layer of SiO2 onto the mesh/PDMS and exposure of the silicone target substrate (Ecoflex 0030, Smooth-On, Inc.) to UV–ozone (to create reactive –OH groups on the surface) enables bonding between the two upon physical contact [21].(Low pressures avoid contact between the PDMS and the finger-tube, thereby allowing bonding only to the mesh.) The SiO2 adhesion layer does not serve any electronic function. Removal of the stamp completes the transfer process, as shown in figure 1(c).
The electrotactile electrodes use 600 nm thick layers of Au in a concentric design, consisting of an inner disk (400 µm radius) surrounded by an outer ring (1000 µm radius) with a 250 µm wide gap between the two. The interconnects consist of 100 µm wide traces of Au in serpentine geometries (radii of curvature ∼800 µm); these traces connect the electrotactile electrodes to Si NM diodes (lateral dimensions of 225 µm x 100 µm and thicknesses of 300 nm). Two layers of Au interconnects (200 and 600 nm thick), isolated by a 1.25 µm PI layer and connected through etched PI vias, establish a compact wiring scheme with overlying interconnects. The 600 nm thick Au interconnect layer allows robust electronic contact though the PI vias. The strain gauge arrays consist of four Si NMs (strips with lateral dimensions of 1 mm x 50 µm and thicknesses of 300 nm) electrically connected by 200 nm thick, 60 µm wide Au traces patterned in serpentine shapes (radii of curvature ∼400 µm). The tactile sensors use 200 nm thick Au electrodes and interconnects in the geometry of the electrotactile arrays but with the concentric electrode pairs replaced by single, disk-shaped electrodes (radii ∼1000 µm).
The Ecoflex substrates, which we refer to as finger tubes, adopt three-dimensional forms specifically matched to those of fingers on a plastic model of the hand. The fabrication involves pouring a polymer precursor to Ecoflex onto a finger of the model and curing at room temperature for 1 h, to create a conformal sheet with ∼125 µm thickness. Pouring a second coating of precursor onto this sheet and curing for an additional 1 h doubles the thickness; repeating this process four times results in a thickness of ∼500 µm.
Removing the Ecoflex from the model and completing the cure by heating at 70°C for 2 h forms a free standing structure, i.e. a finger-tube, like the one illustrated in figure 2. Ecoflex is an attractive material for this purpose because it has a low modulus (∼60 kPa) and large fracture strain (∼900 percent). The former allows soft, intimate contact with the skin; the latter enables the ‘flipping-over’ process referred to previously, and described in quantitative detail in section 3. Transfer printing delivers the device mesh structure to the outer surface of the finger-tube, while pressed into a flattened geometry (figure 2(b)). The entire integrated system is then flipped inside out, to move the mesh from the outer to the inner surface of the tube, as shown in figures 2(c) and (d). Multifunctional devices incorporate electrotactile stimulators on the inside, and strain gauge arrays and tactile sensors on the outside.
3. Results and discussion
The device designs described previously have the advantage that they are conformal to the finger, in a way that naturally presses the electronics on the interior surface of the fingertube (in this case the electrotactile stimulating electrodes) into intimate contact with the skin. The flipping-over process represents a critical step, enabled by careful design of the mechanics in the device mesh. Quantitative mechanics modeling provides important insights. The finger-tube can be approximated as a self-equilibrated, axisymmetric tube with two-dimensional symmetry. Energy minimization using linear elastic shell theory determines the resulting shapes.
Figure 3(a) shows analytical and FEM results for an Ecoflex cylinder with a radius (Rradial) of 7.5 mm and a thickness of 500 µm when bent back on itself, at a midway point during the flipping-over process. The minimum axial radius of curvature (Raxial) of 596 µm, as indicated in figure 3(a), defines the location of maximum induced strain as the tube is flipped over.
The maximum strains on the inner and outer surfaces in this configuration, as shown in the color map of figure 3(b), are ∼ 30–40 percent. The device mesh structures must, therefore, be able to accommodate strains in this range. This requirement is non-trivial for systems like the ones described here, due to their incorporation of brittle materials such as silicon (fracture strain ∼1 percent).
Results and discussion
Circuit layouts, guided by theory, can be identified to satisfy these requirements. As an example, figure 3(c) provides a diagram of a multiplexed electrotactile array in a mesh configuration with narrow, serpentine interconnects. The orange and blue regions correspond to Au layers separated by layers of PI, respectively; the red regions indicate Si NM (300 nm thick) diodes in a PIN (p-doped/intrinsic/ndoped) configuration. The short dimensions of the diodes lie parallel to the flipping-over direction, to minimize the strains in the Si during this process. These optimizations lead to maximum calculated strains that are only 0.051 percent, 0.10 percent, and 0.040 percent for the Au, PI, and the Si, respectively (see figure 3(d) and supplementary file available at. The computed position of the NMP also appears in figure 3(d). Since the moduli of the device layers are several orders of magnitude larger than that of Ecoflex, the location of the NMP plane is largely independent of the Ecoflex. Appropriate selection of the thicknesses of the PI layers allows the NMP to be positioned at the location of the Si NMs, thereby minimizing the induced strains in this brittle material [21, 22]. The thicknesses of the Si NM diodes influence the maximum strains that they experience, as shown in the analytical calculations of figure 3(f). A minimum occurs at the thickness that places the NMP at the shortest distance from the Si NM diode (i.e. hNMP). The position of this minimum can also be adjusted by changing the thicknesses of the PI layers, for example. Further reductions in strain can be realized by reducing the lengths of the devices. Implementation of designs that incorporate these considerations and together with the use of interconnects with optimized serpentine layouts ensure robust device behavior throughout the fabrication sequence. For example, figure 3(e) shows negligible change in the I–V characteristics (Agilent 4155C semiconductor parameter analyzer) of a Si NM diode before and after the flipping-over process.
The experimental results demonstrate the expected functionality in the electrotactile arrays. Figure 4(a) shows the perception of touch on a dry human thumb as a function of voltage and frequency, applied between the inner dot and outer ring electrodes (figure 3(c)). The stimulation used a monophasic, square wave with 20 percent duty cycle, generated using a custom setup. The inset provides an image of a device, with connection to external drive electronics via a flexible anisotropic conductive film (ACF). The required voltage for sensation decreases with increasing frequency, consistent with equivalent circuit models of skin impedance that involve resistors and capacitors connected in parallel. The absolute magnitudes of these voltages depend strongly on the skin hydration level, electrode design, and stimulation waveform [23]. Figure 4(b) shows I–V characteristics of an electrotactile electrode pair while in contact with a hydrated human thumb, measured through a multiplexing diode. At high positive voltages, the resistance of the diode is negligible compared to the skin; here, the slope of the I–V characteristics yield an estimate of the resistance of the skin–electrode contact plus the skin. The value (∼40 kΩ) is in a range consistent with measurements using conventional devices [24, 25]. The diode is stable to at least 20 V, corresponding to currents of 0.25 mA, which is sufficient for electrotactile stimulation on the skin and tongue [2, 6, 7].
These diodes enable multiplexed addressing, according to an approach that appears schematically in figure 4(c). Each unit cell consists of one diode and one electrotactile electrode pair. Figure 4(d) presents a table of the inputs required to address each of the six electrotactile channels. For example, channel SDA can be activating by applying a high potential (+5 V) to inputs A and E and a low potential (0 V) to inputs B, C, and D, thereby yielding a +5 V bias across the outer ring (+5 V) and inner ring electrodes (0 V) of this channel. This configuration forward biases the Si NM diode, which results in a stimulation current, as shown in figure 4(b).
Figure 5 shows a set of straight, uniformly doped Si NMs as strain gauges addressed with interconnects in a mesh geometry. The FEM calculations summarized in figure 5 reveal the strain profiles in a 1 x 4 array of gauges (vertical strips; the yellow dashed box in the upper inset highlights an individual device) on Ecoflex, under a uniaxial in-plane strain of 10 percent. These results show that the overall strain is mostly accommodated by changes in the shapes of the serpentine interconnects and, of course, the Ecoflex itself. The Si NM gauges experience strains (about 10-3) that are ten times lower than the applied strain, as shown in the inset in figure 5(a).
(a)
FEM results of the maximum principal strain for a 1 x 4 array of gauges (straight, vertical structures near the top of the serpentine interconnect mesh) due to an overall 10 percent strain applied along the longitudinal (y) direction. The upper inset shows the strains in the gauge highlighted by the yellow dashed box. The lower inset provides an image of a fabricated device with a layout that matches that of the FEM results.The ability to use Si NMs as high performance strain gauges in stretchable forms results from the strong piezoresistance properties of Si, combined with serpentine layouts. These characteristics, taken together, determine the fractional change in resistance per applied strain. The associated effective gauge factor (GFeff) can be related to the intrinsic gauge factor of a silicon gauge, GFSi = ΔR (RεSi), where ΔR is the change in resistance, R is the initial resistance, and εSi is the strain in the silicon, by the following expression: GFeff = GFSi (εSi/εapp), where εapp is the strain applied to the overall, integrated system.
The designs reported here yield values of εSi/εapp that are much smaller than one, specifically to avoid fracture-inducing strains in the Si during fabrication, mounting and use over physiologically relevant ranges of strain. Figure 5(b) shows experimentally measured values of ΔR/R (evaluation at 1 V, using an Agilent 4155C semiconductor parameter analyzer) as a function of εapp, which corresponds to GFeff ∼1. By fitting the experimental and FEM results to figure 5(b), the GFSi is ∼95, consistent with a recent report on Si NM strain gauges, with otherwise similar designs, on flexible sheets of plastic [26]. We emphasize that the device parameters, such as the size of the gauge and the dimensions of the serpentine interconnects, enable engineering control over GFeff, from values as large as GFSi to those that are much smaller, with a correspondingly increased range of strains over which measurements are possible.
Figure 5(c) shows a strain gauge array on a finger-tube located near the knuckle region of the thumb, in straight (I) and bent (II) positions. Upon bending, the gauges experience tensile strain, resulting in an increase in resistance, as shown for three bending cycles in figure 5(d). The relative resistance changes suggest that the strain associated with bending reaches ∼6 percent. As expected, side-to-side motions induce no changes. Figure 5(e) highlights a similar array on a thin sheet of Ecoflex, mounted near the metacarpal region of the thumb. Here, the device adheres to the skin by van der Waals interactions, similar to the mechanisms observed in epidermal electronic systems [13]. The images in figure 5(e) correspond to the thumb in straight (III) and sideways deflected (VI) positions. The changes in resistance for the two gauges on opposite ends of the 1 x 4 array for three side-to-side cycles of motion appear in figure 5(f). For each cycle, the change in resistance of the rightmost gauge indicates compressive strain; the leftmost indicates the corresponding tensile strain. The results suggest that arrays of gauges can be used to identify not only the magnitude but also the type of motion. As a final demonstration, we built a type of tactile (pressure) sensor suitable for integration on the fingertube platform. The devices exploit changes in capacitance associated with opposing electrodes on the inner and outer surfaces of the Ecoflex. Applied pressure decreases the thickness of the Ecoflex, thereby increasing the capacitance of this structure. Here, layouts like those for the electrotactile devices serve as inner electrodes; a mirror image of this array mounted in an aligned configuration on the outer surface defines a collection of parallel plate capacitors with the Ecoflex as the dielectric. An array of such devices on the anterior surface of a model of the hand appears in figure 6(a). Figures 6(b) and (c) show images of the inner and outer electrode arrays. The relative change in capacitance with applied pressure for a representative device appears in figure 6(d) (black symbols). Here, the capacitance was measured (Agilent E4980A LCR meter) as a function of the pressure applied with a series of weights mounted on a platform with a constant contact area, taking care to minimize the effects of parasitic capacitances and to eliminate ground loops.
Approximately linear behavior is observed over the range studied, consistent with a simple mechanical model, ΔC/Co= P (EEcoflex–P), where ΔC is the capacitance change, Co is the initial capacitance, P is the applied pressure, and EEcoflex is the effective Ecoflex modulus. This simple model assumes no electrostriction or strain induced changes in the dielectric constant (figure 6(d), black line). Due to the Poisson effect, the devices also respond to in-plane strains (εapplied), as shown in figure 6(d) (red), consistent with the simple model ΔC/Co = [(EA)system/(EA)/electrodes]vεapplied, where the Poisson’s ratio v is 0.496, and (EA)system and (EA)electrodes are the tensile stiffnesses of the system and the electrodes respectively. This type of technology provides a simple alternative to recently reported devices that offer similar functionality, but on flexible substrates, and based on conductive elastomers, elastomeric dielectrics, or compressible gate dielectrics in organic transistors [14, 16, 18, 27, 28].
(a) Sensors on the anterior of the thumb.
4. Summary and conclusion
The results presented here establish some procedures and design rules for electronics and sensors that can be mounted conformally onto the fingers. Other appendages of the body can be addressed in a similar manner. Furthermore, most of the considerations in the mechanics and fabrication are agnostic to the specific device functionality or mounting locations. As a result, many of these concepts can be applied generally, to other types of system and modes of use. Future challenges include the development of capabilities for wireless power supply and data transfer.
Acknowledgments
This work was carried out in part at the Fredrick Seitz Materials Research Laboratory Central Facilities, University of Illinois, which are partially supported by the US Department of Energy under grants DE-FG02-07ER46453 and DE-FG02-07ER46471. The work was also supported by a National Security Science and Engineering Fellowship and the Center for Nanoscale Chemical Electrical Mechanical Manufacturing Systems at the University of Illinois, which was funded by the National Science Foundation under grant CMMI-0328162. Y H acknowledges support from NSF Grant Nos ECCS-0824129 and OISE-1043143. The authors would like to gratefully acknowledge Yun Soung Kim for helping with the silicon transfer printing, Yizhu Xie for his assistance with the silicon doping, Camille Goudeseune for providing the control circuit and programming for electrotactile stimulation, and Xian Huang for technical discussion.
References
[1] Barfield W, Hendrix C, Bjorneseth O, Kaczmarek K A and Lotens W 1995 Presence-Teleop. Virt. 4 329
[2] Matteau I, Kupers R, Ricciardi E, Pietrini P and Ptito M 2010 Brain Res. Bull. 82 264
[3] Tan H Z, Durlach N I, Reed C M and Rabinowitz W M 1999 Percept. Phychophys. 61 993
[4] Sparks D W, Kuhl P K, Edmonds A E and Gray G P 1978 J. Acoust. Soc. Am. 63 246
[5] Danilov Y P, Tyler M E and Kaczmarek K A 2008 Int. J.Psychophysiol. 69 162
[6] Kaczmarek K A, Webster J G, Bachyrita P and Tompkins W J 1991 IEEE Trans. Biomed. Eng. 38 17
[7] Lozano C A, Kaczmarek K A and Santello M 2009 Somatosens. Mot. Res. 26 50
[8] Warren J P, Bobich L R, Santello M, Sweeney J D and Tillery S I H 2008 IEEE Trans. Neural Syst. Rehabil. 16 410
[9] Bach-y-Rita P, Tyler M E and Kaczmarek K A 2003 Int. J. Hum.-Comput. Int. 15 285
[10] Jones L A and Safter N B 2008 Hum. Factors 50 90
[11] Vuillerme N, Pinsault N, Chenu O, Demongeot J, Payan Y and Danilov Y 2008 Neurosci. Lett. 431 206
[12] Vidal-Verdu F and Hafez M 2007 IEEE Trans. Neural Syst. Rehabil. 15 119
[13] Kim D H et al 2011 Science 333 838
[14] Lipomi D J, Vosgueritchian M, Tee B C, Hellstrom S L, Lee J A, Fox C H and Bao Z 2011 Nature Nanotechnol. 6 788
[15] Rogers J A and Huang Y G 2009 Proc. Natl Acad. Sci. USA 106 10875
[16] Someya T, Sekitani T, Iba S, Kato Y, Kawaguchi H and Sakurai T 2004 Proc. Natl Acad. Sci. USA 101 9966
[17] Rogers J A, Lagally M G and Nuzzo R G 2011 Nature 477 45
[18] Kim D H et al 2011 Nature Mater. 10 316
[19] Meitl M A et al 2006 Nature Mater. 5 33
[20] Yu J and Bulovic V 2007 Appl. Phys. Lett. 91 043102
[21] Kim D H et al 2008 Science 320 507
[22] Rogers J A, Someya T and Huang Y G 2010 Science 327 1603
[23] Kaczmarek K A and Haase S J 2003 IEEE Trans. Neural Syst. Rehabil. 11 9
[24] Woo E J, Hua P, Webster J G, Tompkins W J and Pallasareny R 1992 Med. Biol. Eng. Comput. 30 97
[25] Hua P, Woo E J, Webster J G and Tompkins W J 1993 IEEE Trans. Biomed Eng. 40 335
[26] Won S M et al 2011 IEEE Trans. Electron Devices 58 4074
[27] Someya T et al 2005 Proc. Natl Acad. Sci. USA 102 12321
[28] Takei K et al 2010 Nature Mater. 9 821