Freestanding Functional Structures by Aerosol‐Jet Printing for Stretchable Electronics and Sensing Applications

Modern electronic devices, particularly those intended for wearable or human health monitoring applications, require high levels of flexibility and stretchability. Hence devices, as well as interconnects, need to be capable of retaining functionality even when being mechanically deformed. Most approaches towards achieving this rely on printing or transferring structures onto elastomeric substrates that can withstand stretching. However, the processing involved can often be cumbersome, and the structures themselves tend to suffer from poor fatigue and/or are limited by the mechanical properties of the underlying substrate. Here, we have developed a novel aerosol jet printing technique capable of building fully freestanding functional structures layer by layer, which are robust and reliable upon thousands of stretching cycles. The process involves printing a combination of layers of different materials with the desired functionality, onto a substrate coated with a sacrificial film that is subsequently dissolved to release the printed structure. Using this method, we demonstrate freestanding conductive wires that can be used as stretchable interconnects/electrodes, and that also function as strain‐sensors. Additionally, we show that a freestanding capacitive structure functions as a robust, stretchable humidity sensor, paving the way for the development of other multilayer, multifunctional stretchable devices and sensors.


Introduction
As electronic devices become increasingly indispensable to human beings, modern electronics has evolved from being rigid to being flexible and even stretchable. To meet the challenges arising from the growing demand for wearable devices, epidermal sensors, and humanoid robotic technology, significant progress has been made on developing stretchable electronics such as soft transistors, [1] stretchable batteries, [2] stretchable capacitors, [3] conformal actuators, [4] and various on-skin sensors. [5][6][7][8] For almost all of these types of stretchable electronics, The achievable thickness depends on printing speed, nozzle size, ink concentration, and ink flowrate. Since it is possible to make printable inks out of conductive materials as well as polymers, printed stretchable electronic devices are becoming increasingly popular due to the relative ease and simplicity in the design and fabrication processes. For example, 3D printing techniques have been employed to fabricate freestanding stretchable electrodes. [21] However, in order to make the printed electrode self-supportive, the conductive material still has to be mixed with an elastomeric polymer, to achieve good mechanical property at the expense of lowering the electrical conductivity.
Here, an aerosol-jet printing (AJP) based technique is introduced by which we achieve, for the first time, fully freestanding stretchable conductive wires (FSCW) with robust mechanical properties and good conductivity, thus paving the way for fully printed electronics including wiring. The AJP technique has distinct advantages in terms of fast-prototyping, wide ink compatibility with minimum printed feature size down to 10 µm. [22][23][24] In a typical AJP process ( Figure S1, Supporting Information), material inks are first atomized into aerosol droplets via either pneumatic atomization or ultrasonic atomization, depending on the viscosity of the ink. The droplets containing ink particles are then carried to then deposition head using nitrogen gas flow, and then jetted out from a nozzle onto the desired substrate while being focused using a nitrogen sheath gas. The AJP technique can easily overcome difficulties associated with mixing incompatible inks, for, e.g., mixing of water-based ink and oil-based ink, by mixing them in the aerosol flow path after being atomized from separate atomizers. By tuning the flux, it is also possible to adjust the mixing ratio between inks on demand, [25][26][27] which is extremely convenient for fabrication optimization. In our method, we applied both ink-mixing technology and a multilayered design concept to build a fully printed, stretchable, self-supported freestanding conductive wire, which could be stretched to up to 180% of its original length, with a resistance change of under 150%. The FSCW can therefore be used as both a strain sensor, as well as stretchable conductive electrode/interconnect to be integrated into conformable and stretchable electronic devices. Moreover, with this technique, it was possible to build a fully selfsupported conducting circuit in just a single printing round. We show that this printed structure can function as a tactile sensor. Furthermore, the ability to print multiple layers of different materials means that multiple conductive wires could be stacked up into one freestanding wire, separated by a dielectric polymer to prevent shorting. This feature makes single-wire circuitry possible by reducing the level of design complexity for practical stretching applications. In particular, a double layered FSCW that forms a capacitor geometry, comprising a polymeric layer sandwiched between two layers of silver-based conductive material, is shown to function as a stretchable humidity sensor, which shows good response to humidity changes even while being stretched or in the presence of temperature changes. Compared with previous stretchable humidity sensors, [28][29][30] our devices exhibit robustness, simplicity, and multifunctionality. The printing techniques and materials combinations presented here could therefore have a huge potential for application in robotics, wearable devices, and epidermal sensing.

Results and Discussions
A schematic illustration of how AJP is used to fabricate a freestanding stretchable conductive wire is shown in Figure 1a. Two different inks (see Experimental Section for ink preparation) were used in the printing process: silver nanoparticulate ink Adv. Mater. Technol. 2019, 4, 1900048 Figure 1. a) Schematic illustration of the fabrication process of aerosol-jet printed freestanding stretchable conductive wires (FSCW). b) Silver nanoparticle ink in a glass vial for ultrasonic atomization. c) Polyimide ink in a plastic jar for pneumatic atomization. d) A Y-connector that allows in situ mixing of the ink aerosols generated from the two different atomization methods. e) A photo of a FSCW pattern released from the substrate. f) An SEM image showing the cross-section of the wire. was placed in a glass vial (Figure 1b) for ultrasonic atomization and PI ink was placed in a plastic jar (Figure 1c) for pneumatic atomization. A detailed description of how AJP works in singleink or dual-ink mode can be found in our previously published works, [22,26,27] as well as in Figure S1 in the Supporting Information. Briefly, the fabrication of FSCW comprises three distinct steps, pattern printing, curing, and pattern lift-off ( Figure S2, Supporting Information). A glass slide coated with a thin layer (≈400 nm) of polyvinyl alcohol (PVA) was used as a substrate for temporarily hosting the printed pattern prior to lift-off. The PVA film served as a sacrificial layer to be dissolved at a later stage thus releasing the structures printed on top of it. The first layer of the pattern was printed using PI ink to form a 200 µm wide and 30 µm thick thin line, using three printing passes ( Figure S3 and Movie M1, Supporting Information). Note that the smallest feature size that can be reliably obtained is ≈10 µm. [22] In this work, each printing pattern consisted of multiple connected parallel winding-shaped lines with a small shift of around 20 µm in between each line, to build a wide enough yet structural stretchable wire (exact dimensions of the design are given in Figure S4 in the Supporting Information). Thus, the width of the wire could be controlled by the number of the parallel lines designed in the pattern. The purpose of the multiple passes is to gradually build the width up to 50 µm, while the purpose of printing multiple layers is to control the thickness of the printed lines. In the next printing stage, several further layers of material were printed right on top of the first PI pattern from a mixture of silver and PI, which was achieved by combining the aerosol flow of the individual inks using a pneumatic Y connector ( Figure 1d). By properly tuning the flux ratio and flow speed, and by choosing the appropriate number of printed layers, the top layers of the mixed-ink pattern could achieve good electrical conductivity with reasonable mechanical bonding to the underlying PI pattern. Note that the in situ mixing of the Ag and PI ink aerosols during flow allowed for excellent control over the Ag nanoparticle dispersion, which would not have been otherwise possible from a single ink mixture, while simultaneously ensuring good adhesion with the underlying PI support layer. The as-printed samples were then treated at 130 °C for 12 h to burn away any remaining organic solvent and surfactant that may have been present in the inks. Curing temperature were held below 150 °C as higher temperatures may carbonize the sacrificial PVA layer to a certain degree, which would have affected its dissolvability in water in the subsequent lift-off stage. During the curing stage, the printed structures remained adhered to the rigid glass substrate, and were not found to undergo any significant stress concentration due to thermal mismatch. Samples were then immersed in a petri-dish filled with deionized (DI) water for several hours, allowing the PVA layer to fully dissolve. The required time for dissolving PVA depended on the shape and area of the printed structure which affected how quickly water could penetrate and dissolve the PVA underneath. The as-printed wires could be then lifted from substrates to become fully free-standing and self-supported ( Figure 1e).
To characterize the composition of the FSCWs, cross-sectional scanning electron microscope (SEM) images of the wires were taken. From the cross-sectional view ( Figure 1f) it can be clearly seen that the wire consisted of two distinct layers of materials, where the brighter part is the conductive layer built from a mixture of Ag nanoparticles and PI, while the darker layer is composed with PI only. The thicknesses of the FSCWs could be easily varied by controlling either the number of print passes ( Figure S5, Supporting Information), nozzle sizes, ink flow speeds, or fluxes used in the printing process. The Ag-PI mixed layer provides electrical conductivity in the wire. Meanwhile, the underlying PI layer functions as a mechanically robust support to ensure that the wire can endure a reasonable amount of stress. Additionally, the PI layer served to protect the sacrificial PVA layer from being prematurely dissolved by the water content in the Ag-PI ink mixture. It is difficult to precisely determine the volume ratio of Ag and PI in the mixture due to the printing control being based on ink mist flow rates. To roughly calculate the volume ratio between Ag and PI in the mixed layer, ink collection tests were conducted respectively when only one of the ink jar/vial was filled with ink each time (Details can be found in the Supporting Information). For our most commonly used printing conditions, the mass ratio between Ag and PI is approximately around 50:1, resulting in a calculated volume ratio of around 7:1. The small amount of PI mixed with Ag therefore provide better bonding to hold the Ag nanoparticles ensuring good contact amongst themselves and also with the underlying PI support layer at the bottom, leading to robust electrical conductivity during mechanical deformation. Our experiments showed that mixing of PI with Ag nanoparticles, and/or the lower curing temperature used, lowered the conductivity of the FSCW, as compared with a printed pure Ag wire treated at higher temperature (≈200 °C). However, this can be considered as a trade-off of this method, where some degree of electrical conductivity is sacrificed for greater mechanical robustness of the free-standing structure. In practice, the resistance of the printed wire can be controlled by tuning the flow rate of the PI ink, as shown in Figure 2a. This approach may be useful when fabricating conductive wires with specified values of resistance for use in electronics circuitry. In order to determine whether immersion in water affects the electrical properties of the wire during the liftoff process, the resistance of four different printed wires on glass substrates were monitored when immersed in DI water for up to 6 h. No significant change in resistance was observed across the four samples as shown in Figure 2b.
It has been reported that the winding or serpentine shape we have adopted in our FSCW design has good structural robustness to stretching. [18,23] We therefore conduct extensive electrical conductivity and stretchability tests on a single FSCW. The stretching process was controlled by a linear motor (LinMot), which could provide either manual or programed linear displacements at a minimum step distance of less than 100 µm. The FSCW sample was both electrically and physically connected across two aluminum stubs at each end, which were mounted on holders attached to the linear motor and to the wall, respectively (Figure 2c). Continuous resistance measurements were conducted using a standard four-probe method (Figure 2c). Images of the FSCW at different extensions are presented in Figure 2d. Figure 2e shows the variation of resistance of a typical FSCW as a function of its end-to-end extension, caused due to structural deformation while being stretched. The wire failed due to breakage when extended to 180% of its original length. In comparison, when the FSCW was embedded in a polymer film, it failed at less than 50% extension ( Figure S6, Supporting Information). This demonstrates the advantage of our fabrication process over the standard approach to designing stretchable wires, as the free-standing self-supported nature of our FSCW therefore improved the achievable stretchability of the structure. This observation is in agreement with finite element analysis studies in ref. [18] where it was shown that stretching a meandering structure led to out-of-plane deformation. When the same structure was confined by a polymer encapsulation, this out-of-plane deformation led to failure at lower strain due to regions of high stress concentration along the structure. In the case of our FSCW, within a stretching range of up to 50% extension, the resistance of the FSCW changed almost linearly with extension, till close to the breaking point (>170 %). Below this limit, the almost linear relationship between resistance and extension across several samples ( Figure S6  applications. A maximum resistance change of 140% was observed for an extension of around 170%. At the same time, given that the original resistance of the unstretched wire was only around 80 Ω (estimated equivalent conductivity between 1 and 2 × 10 5 S m −1 ), the wire was still conductive enough when fully stretched to be incorporated in an electric circuit as a robust interconnect, or even as a sensing element. The electrical conductivity if the FSCW depends solely on the volume fraction of the Ag in the conductive layer. Mechanical properties of the FSCW were also tested using a low-load electric screw machine, as shown in the force-over-extension curve in Figure S7 in the Supporting Information. Stretching at low extension (<50 %) required relatively small amount of force, making the wire suitable for incorporation into epidermal or wearable devices. The sensitivity of the stretchable strain sensor is defined as (dR/R 0 )/ (dL/L 0 ) where R is the resistance and L is the length of the sensor, and R 0 and L 0 present the original resistance and length of the device. The FSCWs presented here showed a sensitivity of around 0.75, which is comparable to other stretchable strain sensors such as a carbon nanotube-based sensor with a sensitivity around 0.82 [13] and a silver nanowire-based sensor with sensitivity around 2.5. [31] The behavior of the FSCW during bending when mounted on a substrate or skin is expected to be stable as well, as in this configuration the FSCW would be in tension, which in effect is the mechanical state that is probed by the stretching tests.
Fatigue tests were carried using the linear motor which provided periodic stretching up to 50% original length of the FSCW at a rate of 3 cycles min −1 (approximately 1 mm s −1 ). The resistance of the FSCW in both relaxed and stretched states were found to slightly increase over testing cycles. Beyond a certain point (≈800 cycles), the change became insignificant, which showed that the FSCW had reached a stable and reproducible conducting configuration (Figure 2f). The inset of Figure 2f shows identical change of resistance in response to stretching between the minimum and maximum values across different cycles. When the test was carried on further to several thousands of cycles, the resistance would increase again due to irreversible local damage of the wire, which puts a limit on the lifetime of the FSCW as a reliable strain sensor.
SEM images ( Figure S8, Supporting Information) showed that the cracking caused by stretching might be the reason for the recoverable resistance change, whereby electrical contact was recovered when the shape of the wire returned to its original form. To evaluate the hysteresis of the strain sensor, we monitored the change in resistance over different stretchand-release cycles, as shown in Figure S9 in the Supporting Information. It is found that in the operating range of interest, the sensor exhibits good stability, even after repeated cycling. However, extensive fatigue testing sometimes resulted in part of the conductive material to peel away from the supporting layer, resulting in irreversible resistance changes ( Figure S8, Supporting Information). To overcome this problem, we were able to use the AJP for in situ layer-by-layer printing of different functional layers within a single wire platform, by taking advantage of the relatively large distance between the printer nozzle and the substrate (up to 3-4 mm). Figure 3 shows the characterizations of four printed wires with different numbers of layers composited for PI and Ag/PI mixture. In Figure 3b- (1)(2)(3)(4) are SEM images showing the cross-section of the wire, where on the cutting surface, the brighter parts present for the Ag mixture layers and the darker parts present for the PI layers.
Adv. Mater. Technol. 2019, 4, 1900048   Figure 3d-(1-4) shows the top view of the printed wires at one end, indicating how the electrodes are arranged for the conductive layers in each design and how top PI layers are protecting the conductive layers in Figure 3d-2 and d-3. Noticeably, when protecting layer of PI was printed (Figure 3b-2), the wire's resistance became far less sensitive to stretching, yet yielded a superb robustness during fatigue test. The wire showed in Figure 3b-2 generated less than 20% increasing in resistance after about 75 000 cycles of repeated stretching at 100% original length ( Figure S10, Supporting Information).
A free-standing stretchable tactile sensor was fabricated via aerosol-jet printing and used as an epidermal device for sensing touch. The device was built with a pair of stretchable wires connected across interdigitated electrodes (Figure 4a,b), used Adv. Mater. Technol. 2019, 4, 1900048   Figure 4. a) Photographs of aerosol-jet printed free-standing stretchable tactile sensor. b) SEM image showing the interdigitated electrode structure as the functional part of the sensor. c,d) The tactile sensor mounted onto a finger, conforming to bending motion of the finger. e) Resistance signals from the tactile sensor showing response to touching of the interdigitate structure by a fingertip, when the device is straight or bent, respectively. f) A two-terminal LED powered through a dual-path FSCW. g) The currents and voltages measured over the LED while being powered with a dual-path FSCW undergoing extension up to 100% of its original length. h-k) LED connected to battery via dFSCW, subjected to various degrees of stretching, while still remaining operational.
to measure the electrical impedance across the electrodes. In response to contact of conductive objects such as fingers on the interdigitated electrode surface, the electrical impedance between the electrodes was found to change. With the connecting wires being stretchable, the device could be easily mounted on bending objects such as finger joints, while still remaining operational, as shown in Figure 4c,d. A demonstration of an as-fabricated touching sensor mounted on a human finger was recorded in Movie M2 in the Supporting Information. Monitored by realtime measurement, the change of impedance between the wires reflected the contact and separation of a different fingertip against the interdigitated electrodes, thereby generating trigger signals from tactile actions (Figure 4e), regardless of the bending state of the finger on which the sensor is mounted.
A major problem that prevents stretchable wires from being used in complex circuitry is that each individual conducting path has to be designed according to its own extension, which means different wires might encounter crossing, entanglement, or interference with each other. As an advantage from our approach, we were able to demonstrate a "dual-path" FSCW (dFSCW) (as in Figure 3b 4 and d4) which could directly power a commercial two-terminal surface-mount light-emitting diode (LED) from a battery, but connected by a single wire (Figure 4f). We demonstrated that, even with the dFSCW undergoing up to 100% extension, the voltage and current measured over the LED was not obviously affected (Figure 4g), due to the relatively low and stable resistance of the conductive path surrounded by PI. The brightness of the LED was also visually monitored in Figure 4h-k and Movie M3 (Supporting Information), showing the lighting of the LED was not affected.
Interestingly, the multilayered structure of the dFSCW with the top conductive layer uncovered (as shown in Figure 3a-3, b-3, and d-3) is essentially that of a capacitor, where two metallic layers (Ag) are separated by a dielectric layer (PI) in between. It has previously been reported that moisturesensitive polymeric dielectric capacitors, particularly PI-based capacitors, can be potentially used as humidity sensors. [32,33] To confirm whether our aerosol-jet printed dFSCW can detect changes in humidity, we monitored the change in capacitance of the dFSCW as a function of humidity. The setup used for this purpose is shown in Figure 5a, where a nominally identical dFSCW as described in the previous section was mounted along a rod with movable holders at each end, to enable the device to be stretched and/or fixed as desired. The two ends of the dFSCW were connected to two testing leads for capacitance measurement. At a nearby position along the rod, a commercial humidity sensor was mounted (Figure 5a), to provide readings about the environmental humidity. The rod was then inserted into a tube reformed from a syringe (Figure 5b). Air was driven at a controlled flow rate through a bubbler filled with DI water to become humidified, which was then connected to the chamber of the syringe from its injection outlet (Figure 5c). The other side of the syringe was left open to allow free air flow. The humidity in the chamber of the syringe could then be changed by adjusting the air flow rate. The capacitance of the device was measured using an impedance analyzer (model 4294A, Agilent Technologies) by sweeping the ac voltage bias frequency between 80 Hz to 50 kHz. A series of measurements were conducted to monitor the change in capacitance (C p ) across different humidity conditions. The results shown in Figure 5d clearly showed that the capacitance of our device increased with relative humidity. The observed changes in capacitance of the device in response to humidity is as a result of the change in dielectric permittivity of the PI layer in the presence of absorbed water. The measured differenced in capacitance values were more significant at lower frequency (e.g., 100 Hz) than at higher frequencies, leading to better sensitivity to humidity changes at lower ac bias frequencies (Figure 5e). However, capacitance changes measured at higher frequency showed even smaller fluctuations across different measurements (smaller error bars).
Importantly, we show that the dFSCW-based humidity sensor can fully function when stretched to different degrees (Figure 5g,h), due to our chosen structural design. Ideally, in order to generate accurate readings, the humidity sensor should be calibrated for different extensions, which can be simultaneously determined by monitoring the resistance of the exposed conductive layers in strain-sensing mode (requiring the use of uncovered dFSCW instead of covered dFSCW). Interestingly, similar structures but with larger scales have shown little change in the capacitance upon stretching. [34] It should be noted that during the stretching process, it is the strain in the PI layer that mostly affects the capacitance of the device. We therefore used finite element analysis to simulate the strain distribution in our structure at around 80% extension, and found that the strain was concentrated mostly along the edges of the structure, leaving the bulk of the structure under little strain ( Figure S11, Supporting Information). A series of measurements of capacitance change for different extensions was conducted. The monitored capacitances were compared in Figure S12 (Supporting Information), showing that within a range of extension between 5% and 35%, the capacitance of the device remained almost the same within the range of measurement frequencies. The tests showed that the effect on capacitance from stretching is reproducible, and could thus be corrected for a given extension by monitoring the strain separately through resistance measurement as already described above. Capacitances over a combination of conditions including 0%-40% stretching with humidity at 56%, 66%, and 76% were measured at both 100 Hz and 10 kHz to map the trend (Figure 5f,g). Our results show that if the extension of the device is maintained within 10% to 40%, the effect on the change in capacitance due to stretching can be neglected. This is an exceptional feature of our FSCW humidity sensor as epidermal electronics are expected to stretch up to 30%, making our device a precise and reliable method to measure humidity in wearable applications. In addition, we found that moderate temperature changes did not lead to significant changes in capacitance ( Figure S13, Supporting Information), and thus our dFSCW-based humidity sensor was found to be stable across a reasonable operating temperature range.

Conclusions
Aerosol-jet printing has been used to fabricate fully freestanding stretchable conductive wires, as well as functional multilayer structures capable of strain and humidity sensing. The functional structures consisting of an elastomeric polymer base layer, with conductive and functional layers on top, were printed layer by layer on a substrate with a sacrificial PVA film. The PVA film was subsequently dissolved in water to release the printed structures thus yielding freestanding self-supported stretchable interconnects and/or functional devices. Devices made by this method showed good electrical stability and stretchability, and are thus well-suited for applications in stretchable electronics, such as in conformal tactile, strain, and humidity sensing. This unique approach can be extended to various combinations of functional materials in multilayered single wire configurations, by simply stacking-up the different layers of materials through a customized AJP technique. Multilayered wires fabricated from such method could be used to form complete closed circuitry using a single wire, while maintaining the stretchability and simplicity of the structure. Since the AJP technique is capable of rapid prototyping from a wide range of inks and ink viscosities, and can also be modified, as we have shown, to mix inks in situ to create functional nanocomposites, it provides an attractive route towards developing stretchable multifunctional sensors for wearable and epidermal applications, and also for tactile applications in robotics.

Experimental Section
PVA Solution Preparation: PVA powder (polyvinyl alcohol, 87%-90% hydrolyzed, average molecular weight 30 000-70 000, Sigma-Aldrich) was added into DI water at 5 wt%. The mixture was then blended using a magnetic stirrer at room temperature for 48 h, till PVA was thoroughly dissolved in water.
Substrate Preparation: Glass slides were cleaned by ethanol and dried. PVA solution was applied onto the cleaned glass and spin-coated for 2 minutes at speed of 500 RPM in a spin-coater (Laurell). After being fully dried, the thickness of the PVA film was about 400 nm. Figure 5. a) A dFSCW and a commercial humidity sensor mounted side by side along a rod, and placed inside b) a syringe chamber for humidity testing. c) The humidity is tuned by controlling the air flow throw a water-filled bubbler to the syringe chamber. d) Measured capacitance of the dFSCW at different humidity levels. e) Capacitance changes as a function of humidity measured at a low frequency (100 Hz) and a high frequency (10 kHz) sources respectively. f,g) Capacitance changes at different levels of extension, under different humidity conditions.