Flexible and stretchable metal oxide nanofiber networks for multimodal and monolithically integrated wearable electronics

Fiber-based electronics enabling lightweight and mechanically flexible/stretchable functions are desirable for numerous e-textile/e-skin optoelectronic applications. These wearable devices require low-cost manufacturing, high reliability, multifunctionality and long-term stability. Here, we report the preparation of representative classes of 3D-inorganic nanofiber network (FN) films by a blow-spinning technique, including semiconducting indium-gallium-zinc oxide (IGZO) and copper oxide, as well as conducting indium-tin oxide and copper metal. Specifically, thin-film transistors based on IGZO FN exhibit negligible performance degradation after one thousand bending cycles and exceptional room-temperature gas sensing performance. Owing to their great stretchability, these metal oxide FNs can be laminated/embedded on/into elastomers, yielding multifunctional single-sensing resistors as well as fully monolithically integrated e-skin devices. These can detect and differentiate multiple stimuli including analytes, light, strain, pressure, temperature, humidity, body movement, and respiratory functions. All of these FN-based devices exhibit excellent sensitivity, response time, and detection limits, making them promising candidates for versatile wearable electronics.

M echanically flexible and optionally stretchable optoelectronic devices integrated with textiles (e-textile) or conformably adaptable to the human body (e-skin) are attracting great interest to create somatosensory systems that are useful for wearable sensors, health monitoring, drug dispensing, and the study of biological function, as well as for technologies that include human-machine interfaces, soft robotics, and augmented reality [1][2][3][4][5][6][7][8] . To enable mechanical stretchability in optoelectronic devices, engineered structures involving percolative charge-transporting networks in elastomers, conductive path designs (wavy, serpentine, mesh) or architectures (wrinkles, bulking) have been employed to accommodate mechanical strain during mechanical stress [9][10][11] . From a materials design perspective, single crystal silicon nanomembranes, metallic (Ag, Au, Cu) nanostructures (nanoparticle, nanowire (NW), nanoflake, etc.), carbonaceous nanomaterials (carbon nanotubes, graphene, amorphous carbon, etc.), and conducting polymers [poly (3,4ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS), polyaniline, etc.] are promising charge-carrying candidates for wearable electronic skins and textiles [12][13][14][15][16][17] . Nevertheless, there is much room for further advances in terms of multi-stimuli sensory capabilities and sensory performance (sensitivity, response time, etc.) as well as fabrication cost 18 . For instance, the transistors and memory elements for semiconducting Si nanomembranes exhibit excellent performance and stability. Furthermore, a certain degree of stretchability can be achieved via design/fabrication of wavy and nanomesh structures 19,20 . However, hightemperature and vacuum-based production of Si nanomembrane is capital-and energy-intensive. In contrast, the large-scale development and commercialization of metal oxide (MO) semiconductors (e.g., indium-gallium-zinc oxide (IGZO), ZnO, SnO 2 ) and conductors (indium-tin oxide (ITO), FTO) is more mature for rigid displays, touch, and sensor applications and highlights the attraction of MOs over other (semi)conductor families in terms of high mobility, optical transparency, and facile solution fabrication in ambient from inexpensive, benign precursors [21][22][23] . However, bulk crystalline/polycrystalline MOs that can sustain maximum fracture levels of~1% are inherently difficult to stretch to the extent required for wearable electronics. Furthermore, the synthesis/processing of MO nanostructures with aspect ratios comparable to those (>1000) of other nanomaterials is challenging, and to date only hydrothermally derived ZnO NWs been explored for stretchable e-skin electronics 24,25 . In recent years, the blow-spinning method has been used for large-scale fabrication of high-aspect-ratio metal/MO nanofibers (TiO 2 , ZrO 2 , Ag, ITO, etc.) for applications in environmental remediation as well as high-temperature sponges and heaters [26][27][28] . However, no studies have addressed the fabrication of semiconducting nanofibers, which in principal could provide broad electronic tunability in microelectronics for logic elements and sensors, not to mention as active elements in wearable and stretchable devices.

Results
Fiber fabrication and characterization. All nanofibers/FN were fabricated by a blow-spinning technique, which, compared with electrospinning, is a more cost-effective and efficient fiber spinning procedure for polymers and polymer-based materials 29,30 . The typical fabrication consists of dissolving the metal salt(s) (100-200 mg mL −1 ) in ethanol, followed by the addition of an appropriate amount of polyvinyl butyral (PVB; 100 mg mL −1 ) and stirring the solution for 2 h. The use of this particular polymer is found to be critical since when combined with the MO salt precursors it enables efficient fiber formation in contrast to other polymers we examined such as polyvinylpyrrolidone and polyethylene oxide. This precursor solution is then pumped into an air brush fixture with a 0.3-mm nozzle and, simultaneously, 150 kPa air is passed into the air brush through a concentric outer nozzle (Fig. 1a). As shown in Supplementary Movie 1, the precursor solution is expelled by the air stream and solidifies on the substrate/target as inorganic-polymer hybrid fibers. This method is highly efficient and scalable, producing 8 × 14 cm 2 area PVBbased fabric in~5 min, and is only equipment-limited ( Fig. 1b and Supplementary Fig. 1). As illustrated in Fig. 1c, d, optical and scanning electron microscopy (SEM) images reveal that neat asspun inorganic-PVB fibers have diameters of 500-900 nm with lengths of several centimeters. Finally, the fabric is annealed using material-dependent optimized conditions to thermolyze the PVB and obtain a fibrous inorganic network (see "Methods" section for details). FNs of different fiber coverages can be obtained by simply changing the deposition time, and C FN s of~0.15,~0.5, and 2.0 µm −1 were used here for specific demonstrations (Supplementary Fig. 2). show representative transmission electron microscopy (TEM) images, selected area electron diffraction (SAED) patterns, and grazing incidence X-ray diffractions (XRDs) of the IGZO, ITO, CuO, and Cu fibers. The IGZO fibers have a smooth surface with diameters of 200-300 nm and exhibit typical diffuse X-ray scattering rings, indicating amorphous character. In contrast, the ITO fibers, connected by numerous crystalline nanoparticles, have an average diameter of~180 nm, while both the CuO and Cu fibers exhibit average diameters of~70 nm. All of the fibers exhibit an impressive average length of~1 cm. The SAED and XRD spectra of the CuO and Cu fibers reveal their crystalline nature. The conductivities of the semiconducting IGZO and CuO fibers are 0.004 S cm −1 and 0.01 S cm −1 , respectively, and the conductivities of the conducting ITO and Cu fibers are 1.0 × 10 3 S cm −1 and 5.9 × 10 6 S cm −1 , respectively ( Supplementary Fig. 5), and remarkably comparable to those of the corresponding bulk materials 6,31,32 . The potential mechanical flexibility of these inorganic fibers was next examined by SEM (Fig. 1i, j and Supplementary Fig. 6), which shows that some fibers have significant curvature (radius~2.5 µm) but without evidence of crack formation.
IGZO FN-based TFTs and gas sensors. Considering the high mechanical flexibility of the present IGZO fibers (vide infra) and potential nanoscopic charge transport confinement in TFTs, a a b c      Fig. 2b and Supplementary Fig. 7. These TFTs exhibit an average carrier mobility (µ FE ) of~1.0 cm 2 V −1 s −1 , no obvious hysteresis, a threshold voltage (V T ) of +20 V, and an on-off current ratio (I on/off ) of 10 4 -10 5 . Note that the effective channel width (W eff ) depends on the fiber number/diameter, which was estimated optically. This performance is somewhat lower than that of solution-processed, dense-film IGZO TFTs, likely reflecting PVBderived carbon contamination and charge traps on the unpassivated fiber surfaces 33,34 . Nevertheless, these FN-based TFTs have substantial semiconductor surface area-to-volume ratios, suitable for highly sensitive gas sensors. Figure 2c shows the I DS -V GS curves for the IGZO TFTs exposed to different gaseous NO 2 concentrations at 25°C, demonstrating significant current decreases upon gas exposure. For these experiments, NO 2 gas having different concentrations was employed using air (50% relative humidity (RH)) as the carrier gas to mimic the actual environmental conditions. Since NO 2 is an oxidant, the adsorption of NO 2 molecules on the fiber surfaces act as acceptors of negative charge carriers, thus reducing the mobile electron density for the n-type IGZO channel 35,36 . The relative change of saturated current (ΔR=R 0 ¼ ΔI=I g ¼ jI 0 À I g j=I g ) vs. NO 2 concentration is plotted in Fig. 2d, where I 0 and I g denote the initial saturated current and that after exposure to the NO 2 (at V DS = V GS = 80 V), respectively, and ΔI is the net current change. The ΔI=I g displays good linearity with increasing NO 2 concentration, and the calculated slope (sensitivity, S G ) is as high as 33.6% ppm −1 for 25°C operation. Figure 2e, f and Supplementary Fig. 8 show the dynamic sensing performance of these devices for NO 2 detection at concentrations ranging from 20 ppb to 20 ppm in air. After exposure to NO 2 , I on decreases rapidly, and recovery occurs without any additional treatments such as vacuum degassing or thermal annealing, typically required (>300°C) for commercial SnO 2 and ZnO sensors 37,38 . The large surface-to-volume ratio, the fibrous network structure, and the weak physisorption of the gas molecules to the fiber surfaces ensures fast adsorption/desorption processes. The operate. Also, previous studies on nanostructured inorganic-based gas sensors require surface decoration and/or UV irradiation, and their sensory performance (working temperature, detection limit, sensitivity, response and recovery time) is far lower than that achieved by the present IGZO FN-based TFTs 35,36,[39][40][41][42][43] (Supplementary Table 1). Thus there is little room left for further optimization of this platform's response/recovery times by additional tuning and investigating structure-sensing behavior and investigating optimum operating conditions. However, a roomtemperature gas sensor is the best choice to minimize energy consumption. Next, flexible IGZO FN (C FN = 0.15 µm −1 )-based TFTs were demonstrated on 1.5-µm ultrathin polyimide (PI) substrates. This polymer was chosen because of its lightweight, solution processability from precursors, considerable mechanical flexibility, and thermal high stability 5 . These top-gate/top-contact devices utilize an ion-gel gate dielectric, poly(vinylidenefluorideco-hexafluoropropylene)/1-ethyl-3-methylimidazolium bis(trifluoromethylsulfo-nyl)-imide (PVDF-HFP/[EMIM][TFSI]; C i = 10.7 µF cm −1 ) and e-beam evaporated Ti/Au (20/100 nm) S/D electrodes (structure in Fig. 2h). The transfer curves acquired under different bending radii are shown in Supplementary Fig. 7, and the calculated mobility/threshold voltages (V T ) are plotted in Fig. 2i. These TFTs function at low voltage (<10 V) and exhibit an average µ FE of 0.52 cm 2 V −1 s −1 and an I on/off of~10 4 . Upon bending the TFT to a 1-mm radius, the µ FE decreases from 0.52 to 0.06 cm 2 V −1 s −1 and V T shifts from +3.7 V to +1.3 V, but both metrics recover to the original values when bending is reversed. Also, the mobility remains constant even after 1000 bending cycles ( Fig. 2j) demonstrating that the IGZO fibers are not damaged by mechanical bending.
IGZO and CuO FN-based devices. Before integrating MO FNs into functioning stretchable devices, the tolerance to mechanical strain was first investigated for an IGZO single fiber and the corresponding FNs (C FN = 0.5 µm −1 ) on SEBS substrates (>600% elongation and tenacity); this was accompanied by a finite element analysis (FEA). Figure 3a, b show optical images of a single IGZO fiber on SEBS before and after application of a 5% tensional strain along the fiber axis, respectively, revealing crack formation (gap = 1.25 µm) at this strain. The crack density and gap distance increase as the strain increases (Supplementary Figs. 9 and 10). Figure 3c and Supplementary Fig. 11 show two crossed IGZO fibers and a bent IGZO fiber under large strains (40-50%). Interestingly, the fiber/part parallel to the stress direction exhibits several cracks while the perpendicular one/part do not break, indicating that fiber orientation has a significant influence on the fiber structural integrity upon stretching. Thus, when considering a randomly oriented FN, those fibers not aligned with the strain direction are much less stressed, which should prevent catastrophic rupture of the conductive pathways under large strains ( Supplementary Fig. 12). To understand this aspect quantitatively, FEA simulations were next carried out. Figure 3d shows the computed relationship of the angle between a fiber long axis vs. the tensional direction (α) and strain (ɛ) exerted on the substrate to achieve a 1.1-GPa stress on the fiber, by assuming the IGZO fiber forms a crack at 0.8% strain 44,45 . Clearly ɛ increases polynominally with α. These data clearly demonstrate that, for achieving the maximum stress (1.1 GPa) on the fiber, a strain of only 0.8% and considerably more, of 22%, must be applied to the surrounding elastomer when α = 0° (Fig. 3e) and 40° (Fig. 3f), respectively.
Since Cr/Au electrodes crack upon large mechanical deformations 46 , to examine the charge transport of the IGZO FN-based resistors under different mechanical strains, the highly stretchable conducting composite PEDOT: PSS)/[EMIM][TFSI] was used as the electrical contacts (L = 100 µm) 47 (Fig. 3g and Supplementary  Fig. 13; Table 1). Note that the entire device array is transparent with an~80% transmittance in the visible ( Supplementary  Fig. 14). The I-V curves of the IGZO FNs with the PEDOT: PSS/[EMIM][TFSI] electrodes were measured under different tensional strains (0-50%, Fig. 3h and Supplementary Fig. 15), and the resulting ΔR/R 0 values are plotted in Fig. 3i. No major resistivity change is observed when the IGZO FN film is stretched up to 25%, at which point ΔR/R 0 increases considerably to 5.1 when stretched to 50%. Importantly, the resistivity returns to near the pristine state after recovery, thus demonstrating the superior stretchability of this IGZO FN. Note that ΔR/R 0 strain simulations based on a randomly oriented IGZO FN exhibit a trend similar to that measured experimentally (Fig. 3i). Furthermore, 5000 stretching cycles (strain 0-10%) do not degrade the electrical response of these IGZO FNs (Fig. 3j).
Owing to the unusual charge transport properties and excellent stretchability of the IGZO FN/SEBS films, they were next utilized to fabricate a multifunctional electronic device for sensing light, chemicals, force, and temperature (Table 1). Since charge transport in amorphous IGZO greatly depends on the oxygen vacancy content, which can be manipulated by photo-induced electron-hole pair formation 48 , the IGZO FN-based resistors were first used for stretchable, wearable monitoring of UV radiation. The electronic properties of these stretchable devices were evaluated under a 0% and 10% strain at a 365 nm wavelength and an intensity of~7.3 mW cm −2 ( Fig. 4a and Supplementary Fig. 16). The photoresponsivity (R ph ) is defined as R ph ¼ ðI ph À I d Þ=AL ph , where I ph and I d are the currents under UV light and in the dark, respectively, A is the effective area of the detector, and L ph is the intensity of the UV light 49 . Another important parameter of a photodetector is the specific detectivity (D*), which is calculated from the equation where q is the absolute value of electron charge and J d the current density under dark 50 . The calculated R ph , on/off current ratio (I ph /I d ), and D* at a voltage of 5 V are shown in Fig. 4b. Note that R ph decreases from 36 to 16 mA W −1 when the device is stretched to 10% strain, while I ph /I d and D* increase from 123/4.4 × 10 10 Jones (no strain) to 403/5.2 × 10 10 Jones (10% strain), owing to the increased effective sensing area achieved on stretching. These metrics rival or exceed those of other MO-based photodetectors 51,52 .
Next, the same IGZO FN resistor was utilized as an artificial nose to detect a poisonous gas (NO 2 ). Figure 4c show the dynamic response to gas exposure, revealing that independent of the degree of strain, this device exhibits excellent response and recovery behavior, which occurs within 5 s. The ΔI=I g for the unstrained device is 7.6, which increases to 33.2 under a 10% strain possibly due to the enlarged sensing area, then it decreases to 9.3 under a 50% strain. At 50% strain, although the sensing area is further increased, the resistance of the IGZO FN-based devices increases greatly due to greater densities of broken fibers. Thus the effective fiber number decreases, resulting in a decreased response. Thus the S G is 38.0% ppm −1 (no strain), 166% ppm −1 (10% strain), and 46.5% ppm −1 (50% strain), respectively. Also, as shown in Fig. 4d and Supplementary Fig. 17, this device is highly selective to NO 2 as verified by control experiments carried out in the presence of other gases-20 ppm NH 3 , 10 4 ppm CO 2 , and 10 3 ppm H 2 . Finally, the function of the same IGZO FN sensor as a temperature sensor/thermistor was investigated over the temperature range 35-75°C. Figure 4e and Supplementary  Fig. 18 show the IGZO FN device resistance dependence (at 5.0 V) on temperature with/without a 10% strain, which clearly indicates a thermally activated electron de-trapping mechanism and a negative temperature coefficient 53 . The resistance exhibits a linear temperature dependence, and the resistance sensitivity (S T ), defined as S T ¼ δðΔR=R 0 Þ=δT 100%, is estimated to be 2.1%°C −1 (no strain) and 2.2%°C −1 (10% strain) via a linear least-squares fitting, which is significantly higher than for previously reported temperature sensors fabricated with IGZO films (0.5%°C −1 ), graphene (oxide)/polymer films (1.4%°C −1 ), and P3HT/polydimethylsiloxane (PDMS) films (1.6%°C −1 ) 6,54-59 (Fig. 4f). Moreover, the present IGZO FN sensors also register response to RH. As shown in dynamic I-t curves ( Supplementary Fig. 19), these devices exhibit good cycling response behavior, with the current increasing with increasing RH since H 2 O adsorption donates electrons to the oxide lattice 60 . Thus, when the IGZO FN devices are cyclically exposed to dry air (RH < 1%) and ambient air (RH = 50%), the relative responsivity ΔI=I 0 is 23% and the response and recovery times are 2 s and 12 s, respectively. Higher ΔI=I 0 (260%) and longer recovery times (120 s) are observed for the device cyclically exposed to 100% RH air and ambient air.
Owing to the excellent gas, temperature, and humidity sensitivities and selectivity of the present IGZO FN resistor (Supplementary Fig. 20 and Supplementary Note 1), it was also applied to human breath analysis, a non-invasive tool to diagnose/monitor respiratory diseases and volatile substance intake 61 . There are two diagnostic abnormal breathing symptoms, bradypnea and tachypnea. Bradypnea, induced by medicines, toxins, head injuries, etc., is abnormally slow breathing, while tachypnea, induced by O 2 deprivation, overheating, anxiety, etc., is associated with abnormally rapid breathing. Figure 4g demonstrates that the IGZO FN sensor outputs a rapid current response (within 2 s) to exhaled breath gas and distinguishable patterns of different respiratory states. When breathing at a slow rate (~3-4 breaths min −1 ), the current recovers to the initial value (~40 nA). In contrast, the current cannot fully recover and remains at a high level (300-6000 nA) for breathing at a normal rate (~15 breaths min −1 ). To mimic tachypnea caused by O 2 deprivation, the person under testing jogged for 5 min, resulting in a respiratory rate of~30 breaths min −1 . In response, the IGZO FN device responds with an~8× increase in saturation current, reaching~5 × 10 4 nA, which is mainly due to the elevated breath temperature and H 2 O vapor after exercise. The current change is only 2-3× the minimum inhaling current of~1.6 × 10 4 nA during rapid breathing. In short, through monitoring current changes, IGZO FN devices can distinguish different breathing rates, elevated temperature, and humidity in exhaled gas after physical activity.
In addition, the wearable IGZO FN devices can also detect human alcohol consumption (right part of Fig. 4g). When testing at a slow breathing rate, the saturation current after drinking 200 mL beer decreases by 50% vs. that without alcohol, and the current change during inhaling and exhaling is 10-20× lower  than that without alcohol (~200×). The mechanism is likely similar to that of water due to the similarity in the chemical structures (H-O-H vs. Et-O-H), with a decrease of the resistance upon surface exposure. The reduction is lower than for water probably because the acidity of water is greater than that of ethanol. Note that small peaks appear between the inhaling/ exhaling breathing cycles likely due to desorption/evaporation of the alcohol molecules. When breathing normally, such small peaks disappear in the next cycle, exhaling coming before the alcohol desorption.
Since the present CuO fibers exhibit a higher Young's modulus (200-300 GPa) than that of IGZO fibers (~110 GPa), this should endow CuO FN-based devices with a higher sensitivity to mechanical strain 6,62 . Therefore, the CuO NFs (C FN = 0.5 µm −1 ) were employed to fabricate a resistor highly sensitive to pressure (Fig. 4h). It is expected that the contacts between the loosely packed nanofibers will increase with increasing pressure, thus increasing the conductivity 63 . The current response to different pressures (0-20 kPa) is shown in Fig. 4h and Supplementary Fig. 21, indicating that this device is sensitive to pressures < 4 kPa with a sensitivity (S p ) of 0.04 kPa −1 . Also, this device can sense the very small pressure exerted by a dime (pressure~50 Pa) and exhibits very short response/recovery times (40/60 ms) when a finger touches/leaves the CuO FN/SEBS device (Fig. 4i, j and Supplementary Fig. 22). Note that we attempted to fabricate TFTs using CuO FNs; however, these devices could not be switched off, which is typical of p-type oxides 64 .
Multifunctional and wearable FN-based e-skins. Using the present blow-spinning method, free-standing IGZO, ITO, and CuO FN films (C FN~2 .0 µm −1 , size = 1 × 2 cm 2 ) were produced that can be laminated directly onto SEBS substrates and contacted with Cu leads (Fig. 5a and Supplementary Fig. 23, Table 1). Owing to the greater conductivity of ITO vs. IGZO and CuO FN films (see Table 1), ITO FN/SEBS devices were used for gesture recognition, in which large strains (up to 20%) occur near one's hand joints 13 . The resistance of the ITO FN film at rest is 7.7 MΩ, and the relative resistance change was measured for strains from 0% to 50% (Fig. 5b). The resistance of the ITO FN/SEBS device increases gradually with strain but returns to the initial value after releasing the strain. The mechanism of resistance change upon strain is similar to that mentioned for the IGZO FN-based devices above. Also, this device is quite stable with similar ΔR/R 0 strain behavior even after storage in ambient for 1 year (Fig. 5b). Next, the resistance change of this platform under different degrees of finger movements, which correspond to different extents of bending angle/strain, was evaluated. The electrical resistance is 7.  b Calculated photoresponsivity, on/off current ratio, and resistor detectivity under no strain and 10% strain. c Cycling test for an IGZO FN/SEBS sensor exposed to 20 ppm NO 2 under no strain, 10%, and 50% strain (bias = 5 V). d Relative resistance changes of IGZO FN/SEBS sensor exposed to 20 ppm NH 3 , 20 ppm NO 2 , 1000 ppm H 2 , and 10,000 ppm CO 2 . e Resistance change of IGZO FN/SEBS device over a 35-75°C temperature range under no stain and 10% strain. f Comparison of temperature response sensitivity with literature reports. g Current change of a wearable resistor exposed to exhaled gas during respiration. Inset: enlarged I-t curves for slow breath rate after drinking alcohol. h Relative current of a CuO FN/ SEBS pressure sensor exposed to different pressures. Inset: schematic of a CuO FN/SEBS pressure sensor. i Dynamic I-t curve of pressure sensor by loading and unloading a dime covered by a piece of tape. Inset: schematic of a dime on CuO FN/SEBS pressure sensor. j Enlarged I-t curve when a finger contacts and disengages the sensor. Inset: enlarged plots show the response and recovery time.
strain), 30°(ɛ~8%), 60°(ɛ~13%), and 90°(ɛ~15%), respectively ( Fig. 5c and Supplementary Fig. 24). The cycling tests in Fig. 5d and Supplementary Movie 2 demonstrate the excellent operational stability under hand motion, and response/recovery times are as short as 2/7 s. Next, the direct correlation between the ΔR/R 0 of the strain sensor with specific gestures was monitored by mounting five strain sensors on the finger joints of a hand. For example, to express the sign of "OK", the thumb and index fingers are completely folded (providing a ΔR/R 0 , >2.0), whereas the middle, ring, and "pinky" fingers are straightened (showing a ΔR/R 0 , <0.2). Figure 5e shows a set of gestures, representing "OK," "LOVE," "HI," "GREAT," and "VICTORY," which was translated by configuring the hand into specific corresponding gestures. Finally, a fully integrated wearable e-skin device comprising self-standing IGZO, CuO, and ITO FNs monolithically integrated on the same platform was laminated onto the hand of a person in the form of a band patch (Supplementary Fig. 25). Figure 5f, j show real-time signal recording of different stimuli, solar light exposure, temperature, strain, and exhaled gas. The procedure involves the subject person standing in an air-conditioned laboratory office (~22°C) (I), exiting the office into sunlight (II), walking in sunlight for several minutes (III), entering an airconditioned coffee shop (IV), placing a hot cup of coffee near the sensor (V), then removing it (VI), bending the hand (VII), and then placing the sensor near a hot cup of coffee (VIII), leaving the table and extending the hand (IX), placing it near the mouth and exhaling on it five times (X), and next removing it (XI), then pressing the hand on a table (XII), and finally positioning it at rest (XIII). From these data and the previous discussion, the IGZO FN sensor can efficiently differentiate solar light, temperature, stretching, and breath stimuli. Specifically, solar light has slight positive effect on the conductivity, while increasing temperature increases it substantially. Breath leads to typical cycling changes while stretching decreases the conductivity. However, this IGZO FN device alone cannot distinguish temperature-involved stimuli (activities III, V, and VIII), while these are effectively detected by the CuO and ITO sensors (see ΔI/I 0 variations in Fig. 5g). Regarding differentiating among stimuli II and X (or among V, VII, and XII), while the ITO device cannot achieve that, the CuO and IGZO sensors perform well. In short, the integrated MO NW-based e-skin platform enables a human-machine interface that differentiates among multiple human stimuli with high recognition levels.

Discussion
In this study, we demonstrate the efficient production at scale of several MO and metal nanofibers and FNs using a simple blowspinning technique employing formulations of metal salt(s) and a polymer. The MO FNs were integrated with elastomeric substrates to achieve a new class of FN-based semiconductor and conductor platforms for inorganic electronics and sensors that are highly stretchable without the need for additional mechanical structure designs required for traditional nonstretchable inorganic materials. The enhanced FN film mechanical properties can be understood by computational and experimental analyses showing the capacity of these systems to sustain greater strain reflects the diverse fiber directionalities in the 3D FNs. The present results show that flexible transistors remain functional when bent with a radius of 5 mm for 1000 times and that stretchable resistors function well after 50% stretching and long-time storage. Wearable sensors operate in a very stable manner while achieving remarkable performance in sensing various combinations of strain, gas, light, temperature, pressure, and exhaled gases. The present monolithically integrated MO NW-based sensor system enables human-machine interactions with multiple sensing functionalities and high levels of recognition. We foresee that this new strategy of fabricating elastomeric inorganic semiconductors/ conductors will advance the development of very stable and sensitive stretchable electronic functionality. Furthermore, the integration demonstrated will promote advances in stretchable electronics for a wide range of applications, such as artificial skins, biomedical implants, and surgical gloves. Fiber fabrication. n ++ -Silicon wafers with 300 nm SiO 2 were cleaned with acetone and isopropyl alcohol and then treated with an oxygen plasma for 5 min before use. The precursor solution was loaded into the spray gun with a 0.3-mm nozzle and sprayed on the substrate/aluminum foil placed at a distance of 20 cm using a flow rate of 0.1 mL min −1 and a gas pressure is 100 kPa. The PVB precursor fibers were collected on SiO 2 /Si substrates or aluminum foils, then the salts/PVB FNs were annealed at optimized temperature and atmosphere. Specifically, the IGZO precursor/PVB fibers were annealed in a furnace heated from room temperature (RT) to 450°C in air at a heating rate of 10°C min −1 and then held at 450°C in air for 15-30 min followed by naturally cooling down in the furnace (refer to the DSC/ TGA results in Supplementary Fig. 26). The ITO precursor/PVB fibers were annealed from RT to 500°C in air at a heating rate of 10°C min −1 and then held at 500°C in air for 1 h before being cooled down in the furnace. To increase the conductivity, the resulting ITO fibers were further annealed from RT to 300°C in 5%H 2 /N 2 at a heating rate of 2°C min −1 and then held at 300°C in air for 1 h before being cooled down. The CuO precursor/PVB fibers were annealed from RT to 450°C in air at a heating rate of 10°C min −1 and then held at 450°C in air for 1 h before being cooled down in the furnace. To fabricate Cu fibers, the above CuO fibers were further annealed from RT to 300°C in 5%H 2 /N 2 at a heating rate of 2°C min −1 and then held at 300°C in air for 1 h before being cooled down. The different fiber densities (C FN = 0.15, 0.5, and 2.0 µm −1 ) were achieved by controlling the blow-spun time (30 s, 90 s, and 5 min, respectively).

Methods
Fiber characterization. Optical images were collected by a Leica optical microscope. XRD spectra were acquired with a Rigaku Smartlab workstation using CuKα (1.54 Å) radiation. Thermogravimetric analysis was performed on a SDT Q600 instrument (TA Instruments, Inc.). Experiments were carried out on~2 mg fiber samples and the heating rate of 10°C min −1 under a 70 mL min −1 air flow. TEM and cross-sectional TEM images were collected using a JEOL ARM300F transmission electron microscope. Samples for top-view TEM images were obtained by placing the fibers on water and transferring them onto TEM grids. For the stretchable ITO resistors and systematic integration of IGZO, CuO, and ITO devices, high coverage (C FN = 2.0 µm −1 ) self-standing IGZO, CuO, and ITO FN films were used. First,~1.5 mL SEBS toluene solution (200 mg mL −1 ) was dropcast on a glass slide (2.5 × 7.5 cm 2 ). After 10 min, an appropriate size (e.g., 1 × 2 cm 2 ) of self-standing IGZO, CuO, and ITO FNs were placed directly on the unsolidified SEBS substrate with the aid of a tweezer, followed by positioning two Cu wires (diameter = 0.1 mm) on the FN as external electrodes at a distance of 1.5 cm. During SEBS solidification (store in ambient overnight), the FN and Cu wires penetrate into the SEBS substrates, affording a robust structure.
Electrical characterization. Electrical characterizations were performed with an Agilent 4155C semiconductor parameter analyzer in ambient (RH = 30-40%). For flexible IGZO FN TFTs, the output curve was measured at a gate voltage sweep from 0 to +10 V and drain source voltage from 0 to +1 V. To characterize the transfer curves, each gate voltage sweeps from 0 to +10 V was applied with a source-drain voltage bias of +1 V. The device was mechanically deformed during electrical characterization by laminating the device on different cylindrical objects having a different radius of curvature. For strain tests, the devices were stretched by a custom-made stretcher while measuring their electrical resistance. For gas sensing tests, the IGZO FN devices were stored in an airtight test chamber. To mimic the actual atmosphere, the carrying gas of dry air was purged through saturated aqueous solution of Mg(NO 3 ) 2 to form wet air with RH of~50%. Then a mixture of wet air and certain gas analyte (2 ppm NO 2 /N 2 gas, 100 ppm NO 2 /N 2 gas, 100 ppm NH 3 /N 2 gas, 4% H 2 /N 2 , or CO 2 gas) in appropriate concentrations was introduced into the test chamber by mass flow controllers. The total flow rate for the test was~550 sccm.
For the UV photodetector test, UV light having an intensity of~7.3 mW cm −2 and a wavelength of 365 nm was used to characterize the photodetector response of the devices. The dark current was measured by keeping the device in a black box to simulate a dark ambient. A voltage sweeps from −10 to +10 V was used to check the resistance of device. Further dynamic response of photodetector is observed by switching the UV lamp turning on and off and recording the time response of rise and decay of the photo-generated current in device. The effective illumination area is calculated by multiplying fiber number and diameter. The mechanical deformation was provided by uniaxial stretching of the device. For temperature sensing, a resistive heater was used to increase the temperature environment, which was monitored using a Fluke 572-2 high-temperature infrared thermometer in direct contact with the devices. The current variations of the devices were measured at a fixed voltage bias of 5 V. The mechanical deformation was provided by uniaxially stretching the device. For the wearable breathing gas test, the devices were placed close to the nose while connecting them to the Agilent B1500A semiconductor parameter analyzer. Different breathing frequencies (slow, normal, fast) under different conditions (normal, after jogging, or drinking beer) were recorded.
For humidity sensing, the IGZO FN/SEBS device was exposed to air with three different humidity (1%, 54%, 100%). The current variations of the devices were measured at a fixed voltage bias of 10 V. For pressure sensing, pressure was applied to the sensing area with pressure from 50 Pa to 20 kPa. To test the response and recovery times, a B1500A semiconductor analyzer was employed using the I-t setup supplied with the software. We measured the current changes with a sampling interval of 20 ms at a given voltage when the index finger touched/left the CuO FN/ SEBS device quickly. The finger was covered by a thick tape to eliminate thermal transport.
Regarding the monolithically integrated platform, the stretchable IGZO, CuO, and ITO FN devices were laminated on the hand of person in the form of a band patch. An Xe arc lamp of a Spectra-Nova 300 W Class-A solar simulator was used to simulate solar light, which has an AM 1.5 G irradiation (100 mW cm −2 ). Approaching a hot coffee can elevate the temperature around device platform. Bending the finger leads to an~10% strain and holding the cup of coffee induces strain, pressure, and temperature elevation. Breath analysis is achieved by exhaling gas on the device platform and pressure (~10 kPa) is applied by pressing the table. For the I-t curve of the monolithically integrated platform, baseline corrections were made using the Origin software.
Finite element analysis. The commercial software ABAQUS was utilized to analyze the deformation of FNs embedded in an elastomeric film and to predict the relationship between the tensile strain of the film and the fracture of the fiber with different α. 3D solid elements (C3D10M) were chosen for the heterogeneous elastomeric substrate and fibers. A convergence test of the mesh size was performed to ensure accuracy, where the minimal element size was 1/20 of the diameter of the fiber. The elastic modulus (E) and Poisson's ratio (ν) of inorganic fibers are 110 GPa and 0.34, respectively 44,45 . The crack strain is taken as 0.8%. The elastic modulus (E) and Poisson's ratio (ν) of substrate are 4.8 MPa and 0.5, respectively. A simplified theoretical model is established to qualitatively analyze the relationship between the resistance change and tensile strain of the film. The model assumes that there are n straight fibers evenly embedded in the film as shown in Fig. 3f, where α is the angle between the fiber and the stretching direction, and the mutual interference between fibers is ignored. The total resistance of the film can be expressed as R 0 = R*/2n, where R* is the resistance of a single fiber with length L 0 . If the fibers have α < α* fracture, the ratio of the resistance change can be expressed as (R − R 0 )/R 0 = sinα*/(1 − sinα*). In addition, the FEA results show that the fiber with a larger α can withstand larger tensile strain due to the rotation of the fiber, i.e., the fibers with smaller α will fracture first under stretching.