materials and optimized designs for humanmachine

D10488

www.advmat.de

Vol. 25 • No. 47 • December 17 • 2013

ADMA_25_47_cover.indd 2 11/25/13 5:23 PM

© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 1

www.advmat.dewww.MaterialsViews.com

wileyonlinelibrary.com

CO

MM

UN

ICATIO

N

Materials and Optimized Designs for Human-Machine Interfaces Via Epidermal Electronics

Jae-Woong Jeong , Woon-Hong Yeo , Aadeel Akhtar , James J. S. Norton , Young-Jin Kwack , Shuo Li , Sung-Young Jung , Yewang Su , Woosik Lee , Jing Xia , Huanyu Cheng , Yonggang Huang , Woon-Seop Choi , Timothy Bretl , and John A. Rogers *

Surface electromyography (sEMG) is a non-invasive technique for measuring electrical signals generated by the contraction of skeletal muscles. sEMG involves sensors placed on the skin, and can be exploited for many clinical and research purposes, ranging from diagnosing neuromuscular disorders, [ 1 ] studying muscle pain, [ 2,3 ] and controlling prosthetic or orthotic devices. [ 4 ] In all instances, the measuring devices and, in particular, their interfaces to the skin must maximize signal levels while mini-mizing noise and crosstalk from non-targeted muscle groups. At the same time, form factors that enable soft, conformal adhesion, in ways that do not constrain the motions induced by muscle contractions, are essential for accurate measure-ment and long-term use without irritation or interface failure. Traditional sEMG signal measurements use rigid electrodes coupled to the skin via electrolyte gels, affi xed with adhesive tapes or straps. Although valuable for uses in hospitals and other controlled settings, the ability to extend this type of tech-nology to continuous monitoring in everyday life is frustrated

by discomfort associated with the electrodes, inconveniences that follow from bulk wired connections to external electronics, and reductions in measurement fi delity caused by drying of the gels. These same features also practically limit mounting locations to large, relatively fl at regions of the body, such as the forehead, back, chest, forearm or thigh. The fi ngers, the face, the neck and other areas with sharp curvature, loose skin or high levels of tactile sensitivity cannot be addressed easily.

Recent research into new materials and mechanics designs demonstrates alternative means to integrate electrodes, and even fully integrated electronics and sensor systems, directly with the skin. These platforms exploit hybrid combinations of hard electrode materials with soft supports and matrix poly-mers to yield thin, mechanically compliant devices that can adopt the microscale topology of the skin, and can be mounted on nearly any region of the body. Examples include capacitive, biocompatible electrodes based on electrodes insulated with polydimethylsiloxane (PDMS) [ 5 ] and skin adhesive patches that

DOI: 10.1002/adma.201301921

Prof. J. A. Rogers Department of Materials Science and Engineering Beckman Institute for Advanced Science and Technology, and Frederick Seitz Materials Research Laboratory University of Illinois at Urbana-Champaign Urbana , Illinois , 61801 , USA E-mail: [email protected] Dr. J.-W. Jeong,[+] Dr. W.-H. Yeo,[+] S. Li, W. Lee Department of Materials Science and Engineering Beckman Institute for Advanced Science and Technology and Frederick Seitz Materials Research Laboratory University of Illinois at Urbana-ChampaignUrbana , Illinois , 61801 , USA A. AkhtarMedical Scholars Program, Neuroscience ProgramUniversity of Illinois at Urbana-ChampaignUrbana , Illinois , 61801 , USA J. J. S. NortonNeuroscience ProgramUniversity of Illinois at Urbana-ChampaignUrbana , Illinois , 61801 , USA Prof. T. BretlDepartment of Aerospace EngineeringUniversity of Illinois at Urbana-ChampaignUrbana , Illinois , 61801 , USA Y.-J. Kwack, Prof. W.-S. ChoiSchool of Display EngineeringHoseo University Asan , 336-795 , Republic of Korea

[+]These authors contributed equally to this work.

S.-Y. JungDepartment of Mechanical engineeringPohang University of Science and Technology Pohang , 790-784 , Republic of Korea Y. SuCenter for Mechanics and MaterialsTsinghua UniversityBeijing 100084, China Department of Civil and Environmental EngineeringDepartment of Mechanical EngineeringCenter for Engineering and HealthNorthwestern UniversityEvanston , Illinois , 60208 , USA J. XiaDepartment of Engineering MechanicsTsinghua UniversityBeijing 100084, ChinaDepartment of Civil and Environmental EngineeringDepartment of Mechanical Engineering Center for Engineering and HealthNorthwestern University, Evanston Illinois , 60208 , USA H. Cheng, Prof. Y. HuangDepartment of Civil and Environmental EngineeringDepartment of Mechanical Engineering Center for Engineering and Health Northwestern UniversityEvanston , Illinois , 60208 , USA

Adv. Mater. 2013, DOI: 10.1002/adma.201301921

http://doi.wiley.com/10.1002/adma201301921

Hong

Highlight

Hong

Highlight

Hong

Highlight

2


wileyonlinelibrary.com © 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

CO

MM

UN

ICATI

ON use biomimetic micropillars. [ 6,7 ] A class of technology referred

to generically as epidermal electronic systems (EES) [ 8 ] offers not only electrode interfaces to the skin but also integrated electronics for signal amplifi cation and other functions. The materials and overall layouts of EES yield physical properties – area mass density, thickness and effective mechanical modulus – well matched to the epidermis itself. A consequence is that lamination directly onto the surface of the skin leads to inti-mate conformal contact, through the action of van der Waals adhesion alone. Details of this contact, and its impact on the electrical nature of the biotic/abiotic interface have not, how-ever, been explored in detail. The following presents a series of systematic studies that summarize the key effects, and enable optimization of the performance of EES for sEMG. Examples include measurements obtained from wide ranging areas of the body and use of the resulting data for advanced forms of human-machine interface and control.

Measuring and accurately representing sEMG signals depend strongly on the design of the materials and structures of the sensors. Factors that determine the magnitude of elec-trical noise and crosstalk contamination include the electrode size, the inter-electrode distance, and the electrode shape and layout. For instance, the European concerted sEMG group has determined that electrodes with conventional designs should provide separations of 20 mm and dimensions of 10 mm, measured along the direction of the muscle fi bers, [ 9 ] for optimal results. Geometrical, material, and mechanical effects that infl uence the biotic/abiotic measurement interface of EES for sEMG are fundamentally different, in many respects. The essential issues are examined in detail in representative designs shown in Figure 1 a and 1 b. Both the bar- and disk-type layouts incorporate narrow, thin interconnect wires, and gold elec-trodes (200 nm in thickness) in the form of fi lamentary ser-pentine (FS) meshes (Figure 1 d) for measurement, ground, and reference. The interconnects use encapsulating layers of polyimide (PI; 300 nm in thickness, Sigma-Aldrich, USA) above and below to mitigate bending stresses in the metal; the traces terminate in exposed contact pads at one edge for the purpose of external connection and signal acquisition. The FS mesh electrodes involve bare metal on the bottom surfaces to enable direct electrical coupling to the skin. The layouts yield soft, elastic responses to applied strains, in a manner that both provides conformal contact to the skin and the ability to accommodate natural motions without mechanical constraint or interface delamination. FS traces with broad widths mini-mize contact impedance by providing large areal coverage, but such designs do not minimize principal strains associated with deformation. Increasing the radius of curvature of the FS can mitigate this disadvantage. An optimized combination of width (20 μ m) and radius of curvature (45 μ m) emerge from para-metric studies using the fi nite element method (FEM) (see Sup-porting Information, Figure S1). Simulation and experimental results (Figure 1 d and 1 e) indicate that the resulting electrodes can be stretched over 30% with only 0.94% maximum principal strain in the metals (fracture strain of Au: 1%). This layout ensures robust operation at strain levels well beyond those that can be tolerated by the skin (10 – 20%). [ 10 ] Such EES (see Exper-imental Section for fabrication process) can be integrated onto the skin by direct printing, [ 11 ] or they can be delivered by use of

thin, low modulus silicone membrane substrates mounted on water-soluble polymer sheets (polyvinyl alcohol (PVA), Haining Sprutop Chemical Tech, China). [ 11 ] This PVA sheet serves as a temporary support for manual lamination of the device on the skin, and is subsequently eliminated by applying water. [ 8 ] The result is conformal integration of EES onto the skin, well con-fi gured for sEMG.

Studies of EES that have systematic variations in key para-meters, including the membrane thickness, FS mesh electrode size, and inter-electrode separation, defi ne the dependence of geometry, contact mechanics and layout on signal acquisition. Evaluations involve application of devices with different designs onto a given region of the fl exor carpi radialis of the forearm of a single individual (Figure S2) after gently scrubbing the skin with a 70% isopropyl alcohol swab (Dukal, USA). The setup appears in Figure S2. Placing the forearm securely on a table and squeezing the hand to a well-defi ned bending force at the wrist (20N; Mark-10, USA) yields a reproducible, consistent level of muscle activity. Comparative analysis of different EES designs uses the average of sEMG signals measured during

Figure 1. Designs and modeling results for epidermal electronic sys-tems (EES) for surface electromyography (sEMG), including electrodes for measurement, ground and reference. (a) Image of bar-type designs, (b) Image of disk-type designs. The scale bar is 500 μ m both for (a) and (b). The middle image shows a micrograph of the fi lamentary serpentine (FS) mesh structure used for the electrodes (width = 20 μ m), (c) Table of design parameters. (d) Micrograph of a FS mesh structure before (left) and after (right) applying 30% strain. The inset in the middle illustrates the netural mechanical plane (NMP) position of the metal layers (Au), with coatings of polyimide (PI) above and below. (e) FEM support of the experimental case in (d).

s

d

(a)

w

h

s (b)

Electrode

Bar 1 2 3 Disk 1 2 3

w (mm) 1 0.5 0.25d (mm) 5 2.5 1

h (mm) 10 5 2.5

s (mm) 20 s (mm) 20

(c)

Connection pad

(d)

(e)

AuPI

PI NMP

30% strain

Symmetric structure

100µm

0% 153%

0.94%

0%


3


wileyonlinelibrary.com© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

CO

MM

UN

ICATIO

N

where ρ is the skin electrical resistivity, d is the inter-distance between electrodes, A is the contact area of the EES electrode with the skin, ω is signal source frequency, and ε is the skin die-lectric constant. Measurements indicate that an increase in skin hydration, with other parameters fi xed (i.e., A , d and ω ), reduces Z c (Figure 2 c), consistent with an expectation that hydration level infl uences ρ and ε . [ 15 ] In evaluating the effect of the EES thickness (5, 100, and 500 μ m), measurements include sepa-rate determination of skin hydration level (hydration measure-ment by MoistureMeterSC Compact, Delfi n Inc). Experimental determination of contact impedance involves measurements, using a built-in function of the amplifi er (Model TCP-128BA, The James Long Company, NY) as in, [ 14 ] before and after exfo-liating the stratum corneum by repeated (three times) applica-tion and removal of Scotch TM tape. The results, summarized in Table 1 , indicate a ≈30% decrease in impedance on the fl exor carpi radialis by removal of the stratum corneum. Even after accounting for variations in hydration level (between 19.6 and 20.7 for experiments reported here), the contact impedance increases systematically and signifi cantly as the membrane thickness increases (Figure 2 d). This result is consistent with expected trends in interfacial contact. The EES device that uses a 5 μ m-thick membrane substrate shows remarkably low con-tact impedance, comparable to that of standard gel electrodes (5 μ m-thick membrane EES: 24.4 k Ω @hydration = 19.6 vs. gel electrodes: 16.9 k Ω @hydration = 22.3). In addition to con-tact area, the impedance can depend on the choice of electrode materials. [ 16 ] For realistic applications, an overriding considera-tion is in the use of biocompatible, non-corrosive, electrochemi-cally stable materials that resist surface oxidation. In this sense, Au offers an attractive set of characteristics.

Conformal contact contributes to other aspects of improved performance, such as minimized motion artifacts and back-ground noise. Figure 2 e compares effects of the former in base-line signals obtained from an EES with a 5 μ m-thick membrane and a separate conventional rigid sEMG electrode with coupling gel, located side by side. The ground electrode is shared, to facilitate direct comparison of these two systems. (For measure-ments of motion artifacts, the cutoff frequency of the high pass fi lter was set to 1 Hz.) Skin motions generated by applying peri-odic pressure to the skin at a localized region between the EES and the conventional electrode create parasitic features in the measured signals. Securing all external cabling and wires to the arm and table ensures that these features arise from the elec-trode themselves and their interfaces to the skin. (These features disappear when the electrodes are disconnected from the skin, in otherwise identical setups.) The results show that the EES interface provides signals that are minimally sensitive to skin deformation, by comparison to conventional electrodes. The dif-ferences likely arise from the ability of the EES construction to accomodate motions of the skin without relative motion of the electrodes and skin. By contrast, the gel in a conventional inter-face allows such motion, thereby creating slight time-dependent variations in the impedance and associated signals. Another ben-efi t of EES follows from conformal contact to the skin. As shown in Figure 2 f, the device with a 5 μ m-thick membrane simultane-ously offers the best skin contact and also achieves the smallest background noise (12 μ V RMS ). Improved conformal contact results in increased levels of balance in impedance between the

three consecutive contractions, separated by 5 seconds of rest. An amplifi er (TCP-128BA, James Long Company, NY) with built-in high-pass, low-pass, and notch fi lters and an acquisi-tion system (IOtech Personal Daq 3000, Measurement Com-puting, MA; BCI 2000, National Instruments Data Acquisition Unit) serve as means to record sEMG data. The passband of the data acquisition electronics lies between 10–300 Hz. The low frequency cutoff eliminates any motion artifacts.

Conformal contact of the FS mesh electrodes with the skin is critically important to high fi delity measurement. The inter-face mechanics and soft adhesion at the biotic/abiotic inter-face determine the nature of this contact ( Figures 2 a and 2 b). An analytical mechanics model in which the skin surface morphology is assumed to be sinusoidal with a characteristic amplitude (100 μ m) and wavelength (140 μ m) can capture the essential physics. Here, the critical membrane thickness below which van der Waals forces are suffi ciently strong to drive per-fect, conformal contact can be determined by considering the total energy associated with interfacial contact ( U interface ) as [ 12 ] :

Uinter f ace = UE E S bending + Uskin elas tici ty + Uadhes ion (1)

where U EES_bending , U skin_elasticity , and U adhesion are the bending energy of the EES, the elastic energy of the skin, and the adhesion energy of the contact, respectively (see details of the analytical model in Supporting Information). The energy associated with localized deformations at the edges, which is typically negligible compared to that associated with bending, is not included in the analysis. The EES itself is treated as an effective composite material whose bending energy depends on the bending stiffness of both the EES electrodes (modulus of Au: 97 GPa, thickness including encapsulating PI layers: 800 nm, areal ratio of EES on the membrane: 33.7 %) and the membrane substrate (modulus of the membrane: 69 kPa; Ecofl ex, Smooth-On, USA). [ 11 ] The elastic energy of the skin is mainly determined by the wavelength (140 μ m) and modulus (≈130 kPa) of the skin. [ 13 ] Assuming that the con-tact adhesion is dominated by the interface between the EES, membrane, and the skin, the effective work of adhesion is ≈0.25 N m −1 . [ 11 ]

Conformal contact results when the adhesion energy is larger than the sum of the bending and the elastic energy. The-oretical analysis (Figure 2 b) shows that, for EES as in Figure 1 , the critical membrane thickness is ≈25 μ m. In other words, EES built using membranes with thicknesses smaller than this critical value will establish conformal contact to the skin. The analysis shows good qualitative agreement with experi-mental results (Figure 2 a). In particular, a 5 μ m-thick mem-brane makes excellent conformal contact with a polymer sur-face molded into the shape of human skin (forearm). [ 11 ] At a thicknesses of 36 μ m, air gaps can be observed at the interface; membranes with thicknesses of 100 μ m and 500 μ m show decreasing areas of contact, particularly at wrinkles and pits of the skin surface.

Conformal contact dictates not only the interface adhesion mechanics, but also the electrical contact impedance. This quantity, Z c , can be expressed by [ 14 ] :

| ZC | =Dd

A√

1 + (TDg )2

(2)


4



CO

MM

UN

ICATI

ON

thicker membranes cause increased noise amplitude and fl uc-tuation (Figure S3). All of the following quantitative studies of FS electrode designs use EES with 5 μ m-thick membranes.

three electrode sites, which in turn leads to enhanced common mode rejection at the differential amplifi er. Random variations in contact impedance associated with imperfect contacts using

Figure 2. Colorized scanning electron microscope (SEM) images and measurement results obtained through systematic study of design choices in EES for sEMG. The relevant parameters include thickness of the supporting membrane, as well as electrode shape, size, and inter-electrode distance. (a) Angled and cross-sectional SEM images showing degree of conformal contact between a silicone replica of the surface of the skin (grey) and various thicknesses of elastomer membrane substrates (blue) for EES. (b) Analytical calculation of the energy associated with the interfacial contact (U interface ) between a representative skin surface and a membrane substrate as a function of thickness. Conformal contact occurs when U interface > 0. (c) Contact impedance (Z c ) between EES electrodes and the skin, as a function of level of hydration in the skin, (d) Z c as a function of thicknesses of the membrane substrate for an EES electrode, (e) Measured electrical signal (amplitude) as a function of time during mechanical deformation of the skin, for the case of EES (epidermal) and conventional rigid electrodes with gels, (f) Infl uence of the membrane substrate thickness on background noise of sEMG signals recorded with an EES. (g) Signal-to-noise ratio (SNR) for sEMG measured using EES with electrodes having different shapes and sizes. (h) Signal strength of sEMG measured from the fl exor carpi radialis as a function of inter-electrode distance. (i) Crosstalk signal from the extensor muscle as a function of inter-electrode distance. (j) Crosstalk contamination as a function of the inter-electrode distance.


5



CO

MM

UN

ICATIO

Nnearly any region of the body. Figure 3 shows examples of recordings obtained from the skin of the forearm, face, fore-head, back of the neck and index fi nger. The fi rst demonstra-tion involves EES on both forearms (Figure 3 a). The ground electrode, located in-between the measurement and reference electrodes (20 mm apart at center-to-center distance), deter-mines the common-zero potential (Figure 3 a). Although each bar-type electrode has an areal contact that is roughly half that of a corresponding gel-based conventional electrode used for comparison (Figure S4), the baseline noise levels are compa-rable (Figure S5). Signals obtained during bimanual gestures such as ‘squeeze fi sts’, ‘bend wrists to the left’, ‘bend wrists to the right’, and ‘bend wrists outward’ generate four character-istic sEMG patterns, as shown in Figure 3 b. Figure 3 c illus-trates an example of EES confi gured for measuring sEMG signals on the face (muscle: Masseter ), as an example of a location where conventional gel-based electrode might not be acceptable. Clenching the jaw and moving the mouth create different types of EMG signals, each of which is clearly distin-guishable from the baseline noise (Figure 3 d). Likewise, the EES on the forehead (targeted muscle: Procerus ) and the back of the neck (targeted muscle: Trapezius ) (Figures 3 e and 3 g) yield EMG signals associated with blinking and motions of the neck (Figures 3 f and 3 h), respectively. EES can even be wrapped onto the fi nger (Figure 3 i; index fi nger, targeted muscle: Vin-cula brevia ), where bending creates high-quality sEMG signals, in spite of the relatively small size of the corresponding muscle. The versatility in location demonstrated in these experiments suggests many possibilities in health, wellness monitoring, as well as in human-machine interface (HMI) technologies based on sEMG.

Measurement of sEMG signals on the forearms demon-strates the latter possibility through control of a drone quad-rotor (AR. Drone, Parrot SA, Paris). The process fl ow for HMI in this case appears in Figure S6. Here, EES interfaces record sEMG signals from four different bimanual gestures ( Figure 4 a; EMG signals corresponds to the motions shown in the inset). Classifi cation (see Experimental section for details) involves conversion of the raw data to root-mean-square (RMS) values according to

RMS =√

1

N

∑N

k=1[xk ]2

(3)

where N is the number of samples and x k is the k th sample. These sEMG signals are binned into discrete commands using an linear discriminant analysis (LDA) classifi cation algorithm. The accuracy of this classifi er can then be visualized utilizing a confusion matrix (Figure 4 b), in which the columns and rows represent the predicted class and the actual gestures

The electrical infl uence of geometry in bar- and the disk-type FS mesh electrodes can be determined by examining the signal-to-noise ratio (SNR) of sEMG signals measured using EES devices under controlled conditions. Following recom-mendations of the major standards body for bipolar sEMG electrodes, [ 9 ] the inter-electrode distance is fi xed at 20 mm for these studies. The Table in Figure 1 c summarizes the design para meters, each of which corresponds to a variable shown in Figure 1 a and 1 b. As expected, upon an increase of the elec-trode size, the detected signal amplitudes and SNRs increase, due to an integrative effect on the sEMG signals (Figure 2 g). The same consideration accounts for enhanced signals with disk-type electrodes, which have larger contact areas than cor-responding bar-type designs with similar dimensions perpen-dicular to the muscle fi bers. Among the different shapes and sizes, Bar1 exhibits the best SNR, due likely to an increased probability of signal detection that results from the large elec-trode size in the direction perpendicular to the muscle fi bers.

The inter-electrode distance is a factor that infl uences both the signal amplitude and crosstalk. The latter is important because large signals from surrounding tissues can obscure measurements from targeted muscles. Studies involve the Bar1 design with various inter-electrode distances of 10, 20, 30, 40, and 60 mm. The magnitude of crosstalk is measured on the fl exor carpi radialis from the extensor muscle during isometric contractions. Figures 2 h and 2 i show the mean and standard deviation of the RMS amplitude of the sEMG signals from the target (fl exor carpi radialis) and crosstalk (extensor) muscles as a function of the inter-electrode distance. Although increasing this distance enhances both the signal amplitude and cross-talk, a ratiometric study of crosstalk and target muscle signal (Figure 2 j) demonstrates that the smallest crosstalk contami-nation occurs at a spacing of 20 mm. Increases in crosstalk at larger distances arise from the greater spatial overlap of action potentials at the skin. [ 17 ] Decreasing the electrode spacing from 20mm to 10mm increases the crosstalk contamination to the point that the crosstalk amplitude becomes comparable to that of the target signal. The collective results indicate, then, that the bar-type design with an electrode size of 10 mm × 1 mm and an inter-electrode distance of 20 mm yields the best SNR with the smallest crosstalk in sEMG signals. This conclusion, which is consistent with fi ndings obtained using conventional electrodes, applies only to measurements on the forearm fl exor muscle. Other muscle groups require different dimensions. For example, the optimal design for conventional electrode inter-faces to the tibialis anterior muscle with conventional electrodes involves 10 mm × 1 mm bar type geometries with an inter-elec-trode distance of 10 mm. [ 18 ]

The advanced materials and EES designs described above (Bar1) can be used to acquire high-quality sEMG signals on

Table 1. Contact impedance before and after exfoliation of stratum corneum.

Location Impedance before Exfoliation (k Ω )

Impedance after Exfoliation (k Ω )

Impedance Decrease after Exfoliation (%)

Hydration Level

Right forearm at fl exor carpi radialis 31.6 22.4 29.1 18.2

Right forearm near the wrist 29.5 21.2 28.1 20.6

Left forearm near the wrist 35.4 24.3 31.3 17.7


6



CO

MM

UN

ICATI

ON ‘In’ command is 97.7 % for 43 trials. The overall accuracy of all

four classifi cations is 91.1 %. The most signifi cant confusion is observed for the ‘Out’ command, with 85.5 %. The accuracy can be improved with additional training and experience, enhanced classifi cation algorithms, and the use of additional/different fea-tures in the data. The utility of this last strategy is demonstrated by enhanced classifi cation accuracies observed in bimanual (≈91%) compared to unimanual (≈48 %) gestures. The four gestures control the quadrotor for 5 different motions– 1 st Out: ‘take off’, Right: ‘rotate clockwise’, In: ‘fl y forward’, Left: ‘rotate counter-clockwise’, and 2 nd Out: ‘land’ (Figure 4 c; illustration is in top view). As demonstrated in Figure 4 d, the fl ight of the quadrotor can be successfully controlled in this manner (Movie S1).

In conclusion, the studies reported here establish a set of guidelines in materials, mechanics and geometric designs for EES confi gured to measure sEMG signals. In optimized cases, the van der Waals interface between the EES and the skin yield robust and accurate measurement capabilities, with opportuni-ties that lie outside those afforded by conventional electrode technologies. Many of the same general considerations apply to other measurement modalities, as well as to schemes for elec-trical and/or thermal stimulation.

Experimental Section Fabrication of EES: Spin-coating polydimethylsiloxane (PDMS; 10 μ m

in thickness, Dow Corning, USA) on a glass slide followed by exposure to UV-ozone for 3 minutes creates a surface hydrophilic, for casting a layer of polyimide (PI; 300 nm in thickness through dilution with pyrrolidinone, Sigma-Aldrich, USA). Deposition and photolithographic patterning of Cr/Au (5nm/200nm in thickness) defi nes the electrode and interconnect structure. Encapsulating these features with a layer of PI (300nm in thickness) places the metal at the neutral mechanical plane. The fi nal step involves exposing the EMG electrodes and pads for external connection by patterned reactive ion etching of corresponding regions of the top layer of PI. [ 11 ] The result is a structure with a total thickness of 800 nm in an open, serpentine mesh design. A water-soluble tape (3M, USA) enables retrieval of the device from the PDMS-coated glass substrate and transfer to a thin silicone layer coated on a sheet of PVA. To establish strong bonds to the silicone, layers of Ti/SiO 2 (5nm/100nm) are deposited onto the contacting side of the device. The water-soluble tape is washed away after transferring the device to a silicone layer. Figure S7 (Supporting Information) demonstrates the process of transfer from the glass substrate to a membrane and subsequently to the skin.

Classifi cation of EMG signals for HMI: The classifi cation uses two channels of EMG data collected from each forearm, generated by contractions of the fl exor carpi radialis. A 128-channel high impedance amplifi er (Model TCP-128BA, The James Long Company, NY) yields analog data, converted to digital signals for analysis using MATLAB (The Mathworks, Inc., Natick, MA). To interact with the EMG control system, each participant performs four bimanual gestures: 1) squeeze fi sts, 2) bend wrists to the left, 3) bend wrists to the right, and 4) bend wrists outward. For training purposes, the participants perform each bimanual gesture fi ve times for four seconds each. A baseline trial corresponds to the participant at rest for four seconds. Extraction and real-time classifi cation uses a 250 ms sliding window with 100 ms overlap, to determine four features: root mean square (RMS), waveform length, zero crossings, and slope sign changes as describe elsewhere. [ 19,20 ] A null state corresponds to an RMS value below 105% of the baseline. LDA [ 20 ] serves to classify the features into discrete gestures. Following training, participants perform each gesture three times for three different

(‘actual class’), respectively. As in the fi gure, the four gestures illustrated in Figure 4 a correspond to distinct commands; In, Left, Right, and Out. As an example, the success rate for the

Figure 3. Applications of EES for sEMG signal sensing on various loca-tions of the body. (a) EES on the forearm (target muscle: Flexor carpi radialis ) including a magnifi ed view of the device (inset) where MEA, GND, and REF refer to measurement, ground, and reference electrodes, respectively. (b) sEMG signals recorded from both forearms with four types of gestures. (c) EES on the face (target muscle: Masseter ). (d) sEMG signals associated with clenching the jaw and moving the mouth. (e) EES on the forehead (target muscle: Procerus ). (f) sEMG signals associated with blinking of the eyes. (g) EES on the back of the neck (target muscle: Trapezius ). (h) sEMG signals associated with moving the neck. (i) EES on the index fi nger (target muscle: Vincula brevia ). (j) sEMG signals associ-ated while bending the fi nger.

(a) (b)

(c) (d)

(e) (f)

(g) (h)

(i) (j)

0 5 10 15 20 25-1.0

-0.5

0.0

0.5

1.0

Am

plitu

de (

mV

)Time (sec)

Right Left

RightExtension (right)

Flexion (left)

LeftExtension (left)Flexion (right)

Squeeze outside (both)

Squeeze (both)

0 5 10 15 20 25-0.8

-0.4

0.0

0.4

0.8A

mpl

itude

(m

V)

Time (sec)

Clench(once) (twice)

Mouth to right(once) (twice)

0 5 10 15 20-0.2

-0.1

0.0

0.1

0.2

Am

plitu

de (

mV

)

Time (sec)

Blink

(once) (twice) (3 times) (4 times)

0 10 20 30-0.4

-0.2

0.0

0.2

0.4

Am

plitu

de (

mV

)

Time (sec)

Bend

(once) (twice) (3 times) (4 times)

0 5 10 15 20 25-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

Am

plitu

de (

mV

)

Time (sec)

Move to right(once) (twice)

Nod(once) (twice)

GND REFMEA


7



CO

MM

UN

ICATIO

N

supported by the National Science Foundation under Grant DMI-0328162 and the US Department of Energy, Division of Materials Sciences under Award No. DE-FG02-07ER46471 through the Materials Research Laboratory and Center for Microanalysis of Materials (DE-FG02-07ER46453) at the University of Illinois at Urbana-Champaign. J.A.R. acknowledges a National Security Science and Engineering Faculty Fellowship.

Received: April 29, 2013 Published online:

[1] J. M. Gilchrist , G. M. Sachs , Muscle Nerve 2004 , 29 , 165 . [2] C. Bodere , S. H. Tea , M. A. Giroux-Metges , A. Woda , Pain 2005 , 116 ,

33 . [3] D. Farina , L. Arendt-Nielsen , T. Graven-Nielsen , Clin. Neurophysiol .

2005 , 116 , 1558 . [4] C. Castellini , P. Smagt , Bio. Cyber. 2009 , 100 , 35 . [5] S. M. Lee , J. H. Kim , H. J. Byeon , Y. Y. Choi , K. S. Park , S.-H. Lee , J.

Neur. Eng. 2013 , 10 , 1 .

durations (short, medium, and long). During this testing session, a conservative threshold minimizes other gesture classifi cations that occur during the null state. A priori examination of the training data determines this threshold.

Experiments on human subjects : All experiments were conducted under approval from Institutional Review Board at the University of Illinois at Urbana-Champaign (protocol number: 13229).

Supporting Information Supporting Information is available from the Wiley Online Library or from the author.

Acknowledgements J.-W.J. and W.-H.Y. contributed equally to this work. Authors appreciate the help of Jongwoo Lee with the schematic drawings. This study is

Figure 4. Human-machine interface (HMI) via sEMG signals recorded using EES. (a) sEMG signals recorded on forearms associated with four ges-tures. (b) Confusion matrix that describes the performance of the classifi cation task; the number in the box and the color bar show the number of samples in the classifi cation test. (c) Image of a remote operated quadrotor controlled via EES and the illustration of the control for 5 different motions (top view, not to scale ). (d) Images of quadrotor control by sEMG signals from forearms. Insets show the control gestures.

Flying quadrotor via EES

1 2 3

4 5 6

(a)

(c)

(d)

1) Take off 4) Rotate counter clockwise (180°)

Top view (not in scale)

2) Rotate clockwise (180°) 5) Fly forward

3) Fly forward6) Land

0 5 10 15 20 250.0

0.1

0.2

0.3

0.4

Am

plitu

de (

mV

RM

S)

Time (sec)

Right Left

(b)

Predicted ClassIn Out Left Right

In

Out

Act

ual C

lass

Left

Right 88.9%

0

25

50

5 0 0 40

0 1 46 2

1 47 2 5

42 0 0 1

93.9%

85.5%

97.7%

91.1%


8



CO

MM

UN

ICATI

ON [13] H. Q. Jiang , Y. G. Sun , J. A. Rogers , Y. Huang , Int. J. Solids Struct.

2008 , 45 , 2014 . [14] L.-F. Wang , J.-Q. Liu , B. Yang , C.-S. Yang , IEEE Sensors J. 2012 , 12 ,

2898 . [15] S. D. Campbell , K. K. Kraning , E. G. Schibli , S. T. Momn , J. Inves-

tigat. Dermatol. 1977 , 69 , 290 . [16] W. Besio , A. Prasad , presented at EMBS Ann. Int. Conf. , New York

City, USA , Aug. 30 – Sept. 3, 2006 . [17] K. Roeleveld , D. F. Stegeman , H. M. Vingerhoets , A. Van Oosterom ,

Acta Physiol. Scand. 1997 , 160 , 175 . [18] C. J. De Luca , M. Kuznetsov , L. D. Gilmore , S. H. Roy , J. Biomech.

2012 , 45 , 555 . [19] J. Kim , S. Mastnik , E. André , presented at the 13th int. conf. intel-

ligent user interface , 2008 . [20] T. A. Kuiken , G. Li , B. A. Lock , R. D. Lipschutz , L. A. Miller ,

K. A. Stubblefi eld , K. B. Englehart , J. Am. Med. Assoc. 2009 , 301 .

[6] W. G. Bae , D. Kim , M. K. Kwak , L. Ha , S. M. Kang , K. Y. Suh , Adv. Health. Mater. 2013 , 2 , 109 .

[7] M. K. Kwak , H.-E. Jeong , K. Y. Suh , Adv. Mater. 2011 , 23 , 3949 . [8] D.-H. Kim , N. Lu , R. Ma , Y.-S. Kim , R.-H. Kim , S. Wang , J. Wu ,

S. M. Won , H. Tao , A. Islam , K. J. Yu , T.-I. Kim , R. Chowdhury , M. Ying , L. Xu , M. Li , H.-J. Chung , H. Keum , M. McCormick , P. Liu , Y.-W. Zhang , F. G. Omenetto , Y. Huang , T. Coleman , J. A. Rogers , Science 2011 , 333 , 838 .

[9] H. J. Hermens , B. Freriks , C. Disselhorst-Klug , G. Rau , J. Electro-myogr. Kinesiol. 2000 , 10 , 361 .

[10] N. Lu , C. Lu , S. Yang , J. A. Rogers , Adv. Funct. Mater. 2012 , 22 , 4044 . [11] W.-H. Yeo , Y.-S. Kim , J. Lee , A. Ameen , L. Shi , M. Li , S. Wang , R. Ma ,

S. H. Jin , Z. Kang , Y. Huang , J. A. Rogers , Adv. Mater. 2013 , 25, 2773 .

[12] S. Wang , M. Li , J. Wu , D.-H. Kim , N. Lu , Y. Su , Z. Kang , Y. Huang , J. A. Rogers , J. Appl. Mech. 2012 , 79 , 031022 .


Copyright WILEY-VCH Verlag GmbH & Co. KGaA, 69469 Weinheim, Germany, 2013.

Supporting Information for Adv. Mater., DOI: 10.1002/adma.201301921 Materials and Optimized Designs for Human-Machine Interfaces Via Epidermal Electronics Jae-Woong Jeong, Woon-Hong Yeo, Aadeel Akhtar, James J. S. Norton, Young-Jin Kwack, Shuo Li, Sung-Young Jung, Yewang Su, Woosik Lee, Jing Xia, Huanyu Cheng, Yonggang Huang, Woon-Seop Choi, Timothy Bretl, and John A. Rogers*

Submitted to

1

DOI: 10.1002/adma.201301921

Supporting Information

Materials and optimized designs for human-machine interfaces via epidermal

electronics

By Jae-Woong Jeong, Woon-Hong Yeo, Aadeel Akhtar, James Norton, Young-Jin Kwack,

Shuo Li, Sung-Young Jung, Yewang Su, Woosik Lee, Jing Xia, Huanyu Cheng, Yonggang

Huang, Woon-Seop Choi, Timothy Bretl, and John A. Rogers*

[*] Prof. John A. Rogers Corresponding-Author

Department of Materials Science and Engineering,

Beckman Institute for Advanced Science and Technology, and Frederick Seitz Materials

Research Laboratory, University of Illinois at Urbana-Champaign,

Urbana, Illinois 61801 (USA)

E-mail: [email protected]

Dr. Jae-Woong Jeong[†]

, Dr. Woon-Hong Yeo[†]

, Shuo Li, Woosik Lee

Department of Materials Science and Engineering,

Beckman Institute for Advanced Science and Technology, and Frederick Seitz Materials

Research Laboratory, University of Illinois at Urbana-Champaign,


Aadeel Akhtar

Medical Scholars Program, Neuroscience Program,

University of Illinois at Urbana-Champaign,


James Norton

Neuroscience Program,



Prof. Timothy Bretl

Department of Aerospace Engineering,



Young-Jin Kwack, Prof. Woon-Seop Choi

School of Display Engineering,

Hoseo University,

Asan, 336-795 (Republic of Korea)

Sung-Young Jung

Department of Mechanical engineering,

Pohang University of Science and Technology,

Pohang, 790-784 (Republic of Korea)

Yewang Su

Center for Mechanics and Materials, Tsinghua University, Beijing 100084 (China)

Department of Civil and Environmental Engineering; Department of Mechanical Engineering;

Center for Engineering and Health,

Northwestern University,

Evanston, Illinois 60208 (USA)

Jing Xia

Department of Engineering Mechanics, Tsinghua University, Beijing 100084 (China)

Submitted to

2


Center for Engineering and Health,



Huanyu Cheng, Prof. Yonggang Huang


Center for Engineering and Health



[†] These authors contributed equally to this work.

Keywords: Epidermal electronic systems (EES); Human-machine interface (HMI); Optimized

EES design; Surface electromyography (sEMG)

Submitted to

3

Interface mechanics between EES and the skin

To study the interface mechanics between the EES and the skin, we assume the microscopic

morphology of skin surface as a sinusoidal form [1]

:

1 cos 2 2rough roughy x h x

(S1)

where y(x) is the skin roughness, hrough is the amplitude, and λrough is the wavelength.

We consider the contact status of EES to skin as in either ‘non-conformal contact’ or

‘conformal contact’. In the case of ‘non-conformal contact’, the contacting area is almost

zero, which leads the total energy at the interface zero. But, ‘conformal contact’ case is

regarded as that the EES follows the skin morphology and displacement exactly. When we

consider the displacement of the EES as w(x), the displacement of the skin surface with the

EES is represented as uz(x):

1 cos 2 2roughw x h x

(S2)

1 cos 2 2z rough roughu x y w h h x

(S3)

where h is the maximum deflection of the EES.

Here, the conformal contact can be determined by considering the total energy associated with

interfacial contact (Uinterface) as:

Uinterface = UEES_bending + Uskin_elasticity + Uadhesion (S4)

where UEES_bending, Uskin_elasticity, and Uadhesion are the bending energy of the EES, the elastic

energy of the skin, and the adhesion energy of the contact, respectively.

The bending energy of the EES is given as:

(S5)

(S6)

(S7)

Submitted to

4

, where

(S8)

where α is the areal fraction of the sensor in the EES, Emembrane is the Young’s modulus of the

elastomeric membrane, hmembrane is the thickness, N is the number of layers used in the EES, Ei

is the Young’s modulus of ith

layer, hi is the thickness of ith

layer.

The skin is modeled as a semi-infinite body subject to a surface displacement because it is

much thicker than the EES. Then, the elastic energy of skin is described as:

(S9)

(S10)

where σz is the normal stress on the top surface of skin.[2]

The adhesion energy of interfacial contact is described as:

(S11)

Conformal contact results when the adhesion energy is larger than the sum of the bending and

the elastic energy. The simplified expression is described as:

(S12)

For the skin conditions with 130skinE kPa , 140rough m and hrough ≈ 100µm, the critical

membrane thickness is ~25 µm (Figure 2b).

References

[1] S. Wang, M. Li, J. Wu, D.-H. Kim, N. Lu, Y. Su, Z. Kang, Y. Huang, and J.A. Rogers, J.

Appl. Mech. 2012, 79, 031022.

[2] H. Jiang, Y. Sun, J.A. Rogers, and Y. Huang, Int. J. Solids. Struct. 2008, 45, 2014.

Submitted to

5

(a)

(b)

Radiu

s o

f curv

atu

re (

R)

FEM) unstretched FEM) stretched

0%

0.89%

0%

1.30%

Wid

th o

f serp

entine (

w)

w=10μm

FEM) unstretched FEM) stretched

0%

0.94%

0%

0.85%

0%

0.59%

0%

0.59%

w=15μm

w=20μm

R=45μm

R=35μm

R=25μm

Figure S1. Parametric study of various FS mesh designs by FEM (ABAQUS CAE) (a)

Variation of the radius of curvature, (b) Variation of the FS width. Increasing the radius of

curvature leads to decrease of the maximum principal strain, while wider the FS width

increases the maximum principal strain. By considering the areal coverage, optimal values

for radius of curvature and width of the FS were chosen as R=45μm and w=20μm,

respectively.

Submitted to

6

(a)

Side view Top view

(b)

Target muscle: Flexor carpi radialis

Squeeze

inside

(c)

Target muscle: Extensor

Squeeze

outside

EES EES

Force gauge Force gauge

Right forearm

Figure S2. Experimental method of measuring EMG signals (a) Images of the experimental

setup. EES is attached on the flexor carpi radialis of the right forearm and a bending force at

the wrist is monitored by a force meter (Mark-10, USA), (b) Schematic diagram of flexor

carpi radialis, which is active when ‘squeeze inside’, (c) Schematic diagram of extensor

muscle, which is active when ‘squeeze outside’.

Submitted to

7

(a)

(b)

Electrodes without contact impedance matching

Electrodes with contact impedance matching

0 2 4 6 8 10-0.2

-0.1

0.0

0.1

0.2

Am

plit

ud

e (

mV

)

Time (sec)

0 2 4 6 8 10-0.04

-0.02

0.00

0.02

Am

plit

ud

e (

mV

)

Time (sec)

Impedance @ a hydration level of 20.4

- Measurement = 22.85kΩ

- Reference = 20.8kΩ

Impedance @ a hydration level of 19.1

- Measurement = 121.3kΩ

- Reference = 25.8kΩ

Figure S3. Background noises resulted from electrodes (a) with and (b) without contact

impedance matching.

Submitted to

8

EES (48.1% coverage) Solid electrode

Figure S4. Comparison of areal contact between EES and a solid electrode.

Submitted to

9

5mm

5mm

(a)

(b)

EES (48.1% areal coverage)

Conventional gel-electrode

0 2 4 6 8 10-0.03

-0.02

-0.01

0.00

0.01

0.02

0.03

Am

plit

ud

e (

mV

)

Time (sec)

0 2 4 6 8 10-0.03

-0.02

-0.01

0.00

0.01

0.02

0.03

Am

plit

ud

e (

mV

)

Time (sec)

Figure S5. Comparison of background noises from (a) EES and (b) conventional gel-

electrodes.

Submitted to

10

Body

MotionEMG Sensing

via EES

Signal

Processing

Feature

Extraction

Signal

Classification

Machine

Control via

Wi-Fi

Figure S6. Process flow for EMG-based HMI.

Submitted to

11

EMG sensor

Silicone substrate

Transferred sensor

Silicone substrate

Water-soluble tape

1) Fabricate an EMG sensor 2) Pick up the sensor for transfer

3) Transfer-print the sensor on

thin silicone substrate 4) Laminate the device on skin

PDMS

on glass

Figure S7. Transfer process of EES from a glass substrate to a silicone substrate, and to the

skin.

Submitted to

12

Movie S1.

A movie clip that demonstrates a quadrotor control via epidermal electronics.

materials and optimized designs for humanmachine

Documents